geekradar
Int
Explore GeekRadar

More to explore

Good gear All topics About us Say hello
GeekRadar / The comparison desk

Claude Sonnet 4.5 vs GPT-5

Catalog

Capability & preference

General capability · ECI

A cross-benchmark capability estimate. Index points, not a percentage; intervals show uncertainty.

Claude Sonnet 4.5 Not available
GPT-5 150.0* GPT-5

Human preference · Arena

Overall text preferences from head-to-head votes. Not factual accuracy or a task-success rate.

Claude Sonnet 4.5 1434.6 claude-sonnet-4-5-20250929-high-32k · 95% interval 1431.8–1437.3 · 83,940 votes · 2026-10-15
GPT-5 Not available

Human preference · Arena

Overall text preferences from head-to-head votes. Not factual accuracy or a task-success rate.

Claude Sonnet 4.5 Not available
GPT-5 1406.2 GPT-5 (High) · 95% interval 1401.7–1410.6 · 31,718 votes · 2026-10-28

Model performance

Radar task score: four tests, equal weight. An experimental summary of these tasks, not a universal intelligence rating.

Claude Sonnet 4.5

Radar task score

Unrated

2/4 core tests · not enough evidence

3 benchmark results for this configuration

GPT-5

Radar task score

Unrated

Older results · no current score

7 benchmark results for this configuration

Scientific reasoning

GPQA Diamond · 25%

Expert-written science questions; strict multiple-choice scoring. Chance accuracy is 25%.

Claude Sonnet 4.5 82.3*% Approx. 95% interval: 77.0–87.7% · Older result
GPT-5 86.2*% Approx. 95% interval: 82.1–90.2% · Older result

Novel problem solving

Mystery Game Puzzles · 25%

Best-next-move puzzles in an undisclosed game variant. A narrow reasoning test, not general intelligence.

Claude Sonnet 4.5 Not available
GPT-5 23.0*% Approx. 95% interval: 14.7–31.3%

Advanced mathematics

FrontierMath Tiers 1-3 · 25%

Private v2 mathematics problems with Python tools. Earlier versions and Tier 4 are excluded.

Claude Sonnet 4.5 Not available
GPT-5 55.4*% Approx. 95% interval: 49.7–61.2%

Factual answers

SimpleQA Verified · 25%

Correct factual answers under a guessing prompt, graded by a model. Not a hallucination or abstention score.

Claude Sonnet 4.5 30.7*% Approx. 95% interval: 27.8–33.6%
GPT-5 50.1*% Approx. 95% interval: 47.0–53.2%

End-to-end software

MirrorCode · Separate result

Whole-program reimplementation with hidden tests; ML + Private, two languages. Budget: seven days and ten billion tokens per attempt.

Claude Sonnet 4.5 Not available
GPT-5 4.4*% Approx. 95% interval: 0.0–11.2%

Competition mathematics

OTIS Mock AIME 2024-2025 · Separate result

Competition problems with model-extracted answers checked against an exact key. Earlier grading procedures are excluded.

Claude Sonnet 4.5 77.8*% Approx. 95% interval: 65.5–90.1% · Older result
GPT-5 91.4*% Approx. 95% interval: 84.2–98.6% · Older result

Research mathematics

FrontierMath Tier 4 v2 · Separate result

The hardest private FrontierMath problems, corrected v2 set, with Python tools. Separate from Tiers 1-3 in the core score.

Claude Sonnet 4.5 Not available
GPT-5 22.0*% Approx. 95% interval: 9.1–34.8%

Bars show reported task scores. Whiskers are approximate 95% intervals where standard errors are available. Different reasoning budgets are not equal-cost tests. Missing results are never zero.

Run it locally

Inference · memory estimate

Claude Sonnet 4.5

Closed weights. No supported local deployment or GPU requirement is published.

GPT-5

Closed weights. No supported local deployment or GPU requirement is published.

Claude Sonnet 4.5 vs GPT-5
At a glance
Models

Claude Sonnet 4.5

anthropic
Models

GPT-5

openai
Release date2025-09-292025-08-07
Capabilities
Context window
200,000 tokens
400,000 tokens
Maximum input
200,000 tokens*
272,000 tokens*
Maximum output
64,000 tokens
128,000 tokens
Reasoning
Yes
Yes
Tool calling
Yes
Yes
Structured output
Yes*
Yes*
Open weights
No
No
Knowledge cutoff
2025-07-31
2024-09-30
API mode
Chat*
Chat*
Image input
Yes*
Yes*
PDF input
Yes*
Yes*
Parallel tool callsNot reported
Yes*
Prompt caching
Yes*
Yes*
System messagesNot reported
Yes*
Tool choice
Yes*
Yes*
Web searchNot reported
Yes*
Model license
Proprietary
Proprietary
Benchmark results
  • AI2D_TEST · version unreported · reported score87*Platform unreported Compare results
    Test conditions
    Metric
    reported score*
    Checkpoint version
    Not specified*
  • CC-OCR · version unreported · reported score68.10*Platform unreported Compare results
    Test conditions
    Metric
    reported score*
    Checkpoint version
    Not specified*
  • CharXiv(RQ) · version unreported · reported score67.20*Platform unreported Compare results
    Test conditions
    Metric
    reported score*
    Checkpoint version
    Not specified*
  • EmbSpatialBench · version unreported · reported score71.80*Platform unreported Compare results
    Test conditions
    Metric
    reported score*
    Checkpoint version
    Not specified*
  • HallusionBench · version unreported · reported score59.90*Platform unreported Compare results
    Test conditions
    Metric
    reported score*
    Checkpoint version
    Not specified*
  • Mathvista(mini) · version unreported · reported score79.80*Platform unreported Compare results
    Test conditions
    Metric
    reported score*
    Checkpoint version
    Not specified*
  • MLVU · version unreported · reported score72.80*Platform unreported Compare results
    Test conditions
    Metric
    reported score*
    Checkpoint version
    Not specified*
  • MMBench EN-DEV-v1.1 · 1.1 · reported score88.30*Platform unreported Compare results
    Test conditions
    Metric
    reported score*
    Checkpoint version
    Not specified*
  • MMMU · version unreported · reported score79.60*Platform unreported Compare results
    Test conditions
    Metric
    reported score*
    Checkpoint version
    Not specified*
  • MMMU-Pro · version unreported · reported score68.40*Platform unreported Compare results
    Test conditions
    Metric
    reported score*
    Checkpoint version
    Not specified*
  • OmniDocBench1.5 · 1.5 · reported score85.80*Platform unreported Compare results
    Test conditions
    Metric
    reported score*
    Checkpoint version
    Not specified*
  • RealWorldQA · version unreported · reported score70.30*Platform unreported Compare results
    Test conditions
    Metric
    reported score*
    Checkpoint version
    Not specified*
  • SimpleVQA · version unreported · reported score57.60*Platform unreported Compare results
    Test conditions
    Metric
    reported score*
    Checkpoint version
    Not specified*
  • VideoMME (w sub.) · version unreported · reported score81.10*Platform unreported Compare results
    Test conditions
    Metric
    reported score*
    Checkpoint version
    Not specified*
    Variant
    Subtitles enabled*
  • VideoMME (w/o sub.) · version unreported · reported score75.30*Platform unreported Compare results
    Test conditions
    Metric
    reported score*
    Checkpoint version
    Not specified*
    Variant
    Subtitles disabled*
  • VideoMMMU · version unreported · reported score77.60*Platform unreported Compare results
    Test conditions
    Metric
    reported score*
    Checkpoint version
    Not specified*
  • ZEROBench_sub · version unreported · reported score26.30*Platform unreported Compare results
    Test conditions
    Metric
    reported score*
    Checkpoint version
    Not specified*
Detailed specifications
Specifications 1 · 27 fields
cache creation input token cost
3.75e-6*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-sonnet-4-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
cache creation input token cost above 1hr
6.0e-6*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-sonnet-4-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
cache creation input token cost batches
1.875e-6*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-sonnet-4-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
cache read input token cost
3.0e-7*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-sonnet-4-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
cache read input token cost batches
1.5e-7*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-sonnet-4-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
deprecation date
2026-11-30*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-sonnet-4-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
input cost per token
3.0e-6*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-sonnet-4-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
input cost per token batches
1.5e-6*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-sonnet-4-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
litellm provider
anthropic*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-sonnet-4-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
max input tokens
200,000*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-sonnet-4-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
max output tokens
64,000*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-sonnet-4-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
max tokens
64,000*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-sonnet-4-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
mode
chat*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-sonnet-4-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
output cost per token
1.5e-5*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-sonnet-4-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
output cost per token batches
7.5e-6*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-sonnet-4-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
prompt cache min tokens
1,024*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-sonnet-4-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
search context cost per query
search_context_size_high
0.01*
search_context_size_low
0.01*
search_context_size_medium
0.01*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-sonnet-4-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
supports assistant prefill
Yes*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-sonnet-4-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
supports computer use
Yes*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-sonnet-4-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
supports function calling
Yes*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-sonnet-4-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
supports native structured output
Yes*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-sonnet-4-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
supports pdf input
Yes*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-sonnet-4-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
supports prompt caching
Yes*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-sonnet-4-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
supports reasoning
Yes*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-sonnet-4-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
supports response schema
Yes*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-sonnet-4-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
supports tool choice
Yes*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-sonnet-4-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
supports vision
Yes*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-sonnet-4-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
Specifications 2 · 17 fields
HallusionBench
suite
HallusionBench
workload
reported score
score
59.9*
Scope and conditions
Test conditions
Metric
reported score*
Checkpoint version
Not specified*
OmniDocBench1.5
suite
OmniDocBench1.5
version
1.5
workload
reported score
score
85.8*
Scope and conditions
Test conditions
Metric
reported score*
Checkpoint version
Not specified*
CharXiv(RQ)
suite
CharXiv(RQ)
workload
reported score
score
67.2*
Scope and conditions
Test conditions
Metric
reported score*
Checkpoint version
Not specified*
CC-OCR
suite
CC-OCR
workload
reported score
score
68.1*
Scope and conditions
Test conditions
Metric
reported score*
Checkpoint version
Not specified*
AI2D TEST
suite
AI2D_TEST
workload
reported score
score
87.0*
Scope and conditions
Test conditions
Metric
reported score*
Checkpoint version
Not specified*
EmbSpatialBench
suite
EmbSpatialBench
workload
reported score
score
71.8*
Scope and conditions
Test conditions
Metric
reported score*
Checkpoint version
Not specified*
MMMU
suite
MMMU
workload
reported score
score
79.6*
Scope and conditions
Test conditions
Metric
reported score*
Checkpoint version
Not specified*
VideoMME (w sub.)
suite
VideoMME (w sub.)
workload
reported score
score
81.1*
Scope and conditions
Test conditions
Metric
reported score*
Checkpoint version
Not specified*
Variant
Subtitles enabled*
VideoMME (w/o sub.)
suite
VideoMME (w/o sub.)
workload
reported score
score
75.3*
Scope and conditions
Test conditions
Metric
reported score*
Checkpoint version
Not specified*
Variant
Subtitles disabled*
VideoMMMU
suite
VideoMMMU
workload
reported score
score
77.6*
Scope and conditions
Test conditions
Metric
reported score*
Checkpoint version
Not specified*
MLVU
suite
MLVU
workload
reported score
score
72.8*
Scope and conditions
Test conditions
Metric
reported score*
Checkpoint version
Not specified*
MMMU-Pro
suite
MMMU-Pro
workload
reported score
score
68.4*
Scope and conditions
Test conditions
Metric
reported score*
Checkpoint version
Not specified*
Mathvista(mini)
suite
Mathvista(mini)
workload
reported score
score
79.8*
Scope and conditions
Test conditions
Metric
reported score*
Checkpoint version
Not specified*
ZEROBench sub
suite
ZEROBench_sub
workload
reported score
score
26.3*
Scope and conditions
Test conditions
Metric
reported score*
Checkpoint version
Not specified*
RealWorldQA
suite
RealWorldQA
workload
reported score
score
70.3*
Scope and conditions
Test conditions
Metric
reported score*
Checkpoint version
Not specified*
MMBench EN-DEV-v1.1
suite
MMBench EN-DEV-v1.1
version
1.1
workload
reported score
score
88.3*
Scope and conditions
Test conditions
Metric
reported score*
Checkpoint version
Not specified*
SimpleVQA
suite
SimpleVQA
workload
reported score
score
57.6*
Scope and conditions
Test conditions
Metric
reported score*
Checkpoint version
Not specified*
Specifications · 35 fields
cache read input token cost
1.25e-7*
Scope and conditions
native_context
provider
openai*
mode
chat*
full_native_model_id
gpt-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
cache read input token cost batches
6.25e-8*
Scope and conditions
native_context
provider
openai*
mode
chat*
full_native_model_id
gpt-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
cache read input token cost flex
6.25e-8*
Scope and conditions
native_context
provider
openai*
mode
chat*
full_native_model_id
gpt-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
cache read input token cost priority
2.5e-7*
Scope and conditions
native_context
provider
openai*
mode
chat*
full_native_model_id
gpt-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
input cost per token
1.25e-6*
Scope and conditions
native_context
provider
openai*
mode
chat*
full_native_model_id
gpt-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
input cost per token batches
6.25e-7*
Scope and conditions
native_context
provider
openai*
mode
chat*
full_native_model_id
gpt-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
input cost per token flex
6.25e-7*
Scope and conditions
native_context
provider
openai*
mode
chat*
full_native_model_id
gpt-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
input cost per token priority
2.5e-6*
Scope and conditions
native_context
provider
openai*
mode
chat*
full_native_model_id
gpt-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
litellm provider
openai*
Scope and conditions
native_context
provider
openai*
mode
chat*
full_native_model_id
gpt-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
max input tokens
272,000*
Scope and conditions
native_context
provider
openai*
mode
chat*
full_native_model_id
gpt-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
max output tokens
128,000*
Scope and conditions
native_context
provider
openai*
mode
chat*
full_native_model_id
gpt-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
max tokens
128,000*
Scope and conditions
native_context
provider
openai*
mode
chat*
full_native_model_id
gpt-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
mode
chat*
Scope and conditions
native_context
provider
openai*
mode
chat*
full_native_model_id
gpt-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
output cost per token
1.0e-5*
Scope and conditions
native_context
provider
openai*
mode
chat*
full_native_model_id
gpt-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
output cost per token batches
5.0e-6*
Scope and conditions
native_context
provider
openai*
mode
chat*
full_native_model_id
gpt-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
output cost per token flex
5.0e-6*
Scope and conditions
native_context
provider
openai*
mode
chat*
full_native_model_id
gpt-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
output cost per token priority
2.0e-5*
Scope and conditions
native_context
provider
openai*
mode
chat*
full_native_model_id
gpt-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
search context cost per query
search_context_size_high
0.01*
search_context_size_low
0.01*
search_context_size_medium
0.01*
Scope and conditions
native_context
provider
openai*
mode
chat*
full_native_model_id
gpt-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
supported endpoints
  • /v1/chat/completions*
  • /v1/batch*
  • /v1/responses*
Scope and conditions
native_context
provider
openai*
mode
chat*
full_native_model_id
gpt-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
supported modalities
  • text*
  • image*
Scope and conditions
native_context
provider
openai*
mode
chat*
full_native_model_id
gpt-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
supported output modalities
  • text*
Scope and conditions
native_context
provider
openai*
mode
chat*
full_native_model_id
gpt-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
supports function calling
Yes*
Scope and conditions
native_context
provider
openai*
mode
chat*
full_native_model_id
gpt-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
supports minimal reasoning effort
Yes*
Scope and conditions
native_context
provider
openai*
mode
chat*
full_native_model_id
gpt-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
supports native streaming
Yes*
Scope and conditions
native_context
provider
openai*
mode
chat*
full_native_model_id
gpt-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
supports none reasoning effort
No*
Scope and conditions
native_context
provider
openai*
mode
chat*
full_native_model_id
gpt-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
supports parallel function calling
Yes*
Scope and conditions
native_context
provider
openai*
mode
chat*
full_native_model_id
gpt-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
supports pdf input
Yes*
Scope and conditions
native_context
provider
openai*
mode
chat*
full_native_model_id
gpt-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
supports prompt caching
Yes*
Scope and conditions
native_context
provider
openai*
mode
chat*
full_native_model_id
gpt-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
supports reasoning
Yes*
Scope and conditions
native_context
provider
openai*
mode
chat*
full_native_model_id
gpt-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
supports response schema
Yes*
Scope and conditions
native_context
provider
openai*
mode
chat*
full_native_model_id
gpt-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
supports system messages
Yes*
Scope and conditions
native_context
provider
openai*
mode
chat*
full_native_model_id
gpt-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
supports tool choice
Yes*
Scope and conditions
native_context
provider
openai*
mode
chat*
full_native_model_id
gpt-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
supports vision
Yes*
Scope and conditions
native_context
provider
openai*
mode
chat*
full_native_model_id
gpt-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
supports web search
Yes*
Scope and conditions
native_context
provider
openai*
mode
chat*
full_native_model_id
gpt-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
supports xhigh reasoning effort
No*
Scope and conditions
native_context
provider
openai*
mode
chat*
full_native_model_id
gpt-5*
pricing_scope
Literal source field names and values retained; provider, region, cache/batch/threshold conditions and per-unit meaning must be checked. max_tokens is not automatically a context limit.*
Input modalities
  • text
  • image
  • pdf
  • text
  • image
Output modalities
  • text
  • text

API cost

USD / month

Standard text-token estimate. Excludes taxes, caching, batch discounts, reasoning and multimodal surcharges. Provider limits and regional availability may differ.

Claude Sonnet 4.5

GPT-5

Claude Sonnet 4.5 / Pricing tiers 5
Provider / scopeRateCurrencyAmount / unit
AWS BedrockClaude Sonnet 4.5Batch Global cross-region InputUSD1.5 / 1M tokens
OutputUSD7.5 / 1M tokens
AWS BedrockClaude Sonnet 4.5Standard Global cross-region Cached inputUSD0.3 / 1M tokens
InputUSD3 / 1M tokens
OutputUSD15 / 1M tokens
Google Vertex AIClaude Sonnet 4.5Batch Token tier: 0 – 200,000 Cached inputUSD0.15 / 1M tokens
InputUSD1.5 / 1M tokens
OutputUSD7.5 / 1M tokens
Google Vertex AIClaude Sonnet 4.5Standard Token tier: 0 – 200,000 Cached inputUSD0.3 / 1M tokens
InputUSD3 / 1M tokens
OutputUSD15 / 1M tokens
Google Vertex AIClaude Sonnet 4.5Standard Token tier: 200,000 – No stated upper limit Cached inputUSD0.6 / 1M tokens
InputUSD6 / 1M tokens
OutputUSD22.5 / 1M tokens
GPT-5 / Pricing tiers 4
Provider / scopeRateCurrencyAmount / unit
OpenAIgpt-5Batch Cached inputUSD0.0625 / 1M tokens
InputUSD0.625 / 1M tokens
OutputUSD5 / 1M tokens
OpenAIgpt-5Flex Cached inputUSD0.0625 / 1M tokens
InputUSD0.625 / 1M tokens
OutputUSD5 / 1M tokens
OpenAIgpt-5Priority Cached inputUSD0.25 / 1M tokens
InputUSD2.5 / 1M tokens
OutputUSD20 / 1M tokens
OpenAIgpt-5Standard Cached inputUSD0.125 / 1M tokens
InputUSD1.25 / 1M tokens
OutputUSD10 / 1M tokens

Recorded results

Only the same benchmark, version, workload, scenario and unit share a row. Missing results are not zero. System results include the full submitted hardware and software configuration.

Arena text preference / chinesehuman preference · 2026-10-15 · rating

Claude Sonnet 4.5

1,459.59 rating

Confidence interval: 1,451.42 – 1,467.76

5887 votes

Category
chinese
Evaluated configuration
claude-sonnet-4-5-20250929
Evaluation type
human preference

1,458.46 rating

Confidence interval: 1,450.41 – 1,466.51

5917 votes

Category
chinese
Evaluated configuration
claude-sonnet-4-5-20250929-high-32k
Evaluation type
human preference

GPT-5

1,421.78 rating

Confidence interval: 1,408.89 – 1,434.68

2243 votes

Category
chinese
Evaluated configuration
gpt-5-high
Evaluation type
human preference
Arena text preference / codinghuman preference · 2026-10-15 · rating

Claude Sonnet 4.5

1,488.30 rating

Confidence interval: 1,483.45 – 1,493.14

19950 votes

Category
coding
Evaluated configuration
claude-sonnet-4-5-20250929-high-32k
Evaluation type
human preference

1,484.64 rating

Confidence interval: 1,479.66 – 1,489.63

19782 votes

Category
coding
Evaluated configuration
claude-sonnet-4-5-20250929
Evaluation type
human preference

GPT-5

1,436.04 rating

Confidence interval: 1,428.14 – 1,443.93

6400 votes

Category
coding
Evaluated configuration
gpt-5-high
Evaluation type
human preference
Arena text preference / creative_writinghuman preference · 2026-10-15 · rating

Claude Sonnet 4.5

1,441.86 rating

Confidence interval: 1,435.88 – 1,447.85

12416 votes

Category
creative_writing
Evaluated configuration
claude-sonnet-4-5-20250929
Evaluation type
human preference

1,424.53 rating

Confidence interval: 1,418.65 – 1,430.40

12702 votes

Category
creative_writing
Evaluated configuration
claude-sonnet-4-5-20250929-high-32k
Evaluation type
human preference

GPT-5

1,363.95 rating

Confidence interval: 1,354.46 – 1,373.43

4242 votes

Category
creative_writing
Evaluated configuration
gpt-5-high
Evaluation type
human preference
Arena text preference / englishhuman preference · 2026-10-15 · rating

Claude Sonnet 4.5

1,450.04 rating

Confidence interval: 1,446.10 – 1,453.98

35731 votes

Category
english
Evaluated configuration
claude-sonnet-4-5-20250929
Evaluation type
human preference

1,448.69 rating

Confidence interval: 1,444.85 – 1,452.53

36229 votes

Category
english
Evaluated configuration
claude-sonnet-4-5-20250929-high-32k
Evaluation type
human preference

GPT-5

1,409.08 rating

Confidence interval: 1,403.24 – 1,414.93

13606 votes

Category
english
Evaluated configuration
gpt-5-high
Evaluation type
human preference
Arena text preference / exclude_tieshuman preference · 2026-10-15 · rating

Claude Sonnet 4.5

1,433.86 rating

Confidence interval: 1,430.01 – 1,437.70

60410 votes

Category
exclude_ties
Evaluated configuration
claude-sonnet-4-5-20250929
Evaluation type
human preference

1,427.42 rating

Confidence interval: 1,423.70 – 1,431.15

61729 votes

Category
exclude_ties
Evaluated configuration
claude-sonnet-4-5-20250929-high-32k
Evaluation type
human preference

GPT-5

1,387.01 rating

Confidence interval: 1,381.12 – 1,392.90

23424 votes

Category
exclude_ties
Evaluated configuration
gpt-5-high
Evaluation type
human preference
Arena text preference / experthuman preference · 2026-10-15 · rating

Claude Sonnet 4.5

1,482.54 rating

Confidence interval: 1,474.61 – 1,490.47

6078 votes

Category
expert
Evaluated configuration
claude-sonnet-4-5-20250929-high-32k
Evaluation type
human preference

1,472.00 rating

Confidence interval: 1,464.16 – 1,479.83

6161 votes

Category
expert
Evaluated configuration
claude-sonnet-4-5-20250929
Evaluation type
human preference

GPT-5

1,419.00 rating

Confidence interval: 1,404.19 – 1,433.81

1695 votes

Category
expert
Evaluated configuration
gpt-5-high
Evaluation type
human preference
Arena text preference / frenchhuman preference · 2026-10-15 · rating

Claude Sonnet 4.5

1,458.49 rating

Confidence interval: 1,443.97 – 1,473.00

2001 votes

Category
french
Evaluated configuration
claude-sonnet-4-5-20250929
Evaluation type
human preference

1,450.84 rating

Confidence interval: 1,436.37 – 1,465.31

1990 votes

Category
french
Evaluated configuration
claude-sonnet-4-5-20250929-high-32k
Evaluation type
human preference

GPT-5

1,407.58 rating

Confidence interval: 1,379.08 – 1,436.09

480 votes

Category
french
Evaluated configuration
gpt-5-high
Evaluation type
human preference
Arena text preference / germanhuman preference · 2026-10-15 · rating

Claude Sonnet 4.5

1,426.98 rating

Confidence interval: 1,410.35 – 1,443.61

1355 votes

Category
german
Evaluated configuration
claude-sonnet-4-5-20250929-high-32k
Evaluation type
human preference

1,421.06 rating

Confidence interval: 1,404.59 – 1,437.52

1356 votes

Category
german
Evaluated configuration
claude-sonnet-4-5-20250929
Evaluation type
human preference

GPT-5

1,416.35 rating

Confidence interval: 1,394.11 – 1,438.58

748 votes

Category
german
Evaluated configuration
gpt-5-high
Evaluation type
human preference
Arena text preference / hard_promptshuman preference · 2026-10-15 · rating

Claude Sonnet 4.5

1,461.75 rating

Confidence interval: 1,458.16 – 1,465.34

48326 votes

Category
hard_prompts
Evaluated configuration
claude-sonnet-4-5-20250929
Evaluation type
human preference

1,461.68 rating

Confidence interval: 1,458.15 – 1,465.21

48471 votes

Category
hard_prompts
Evaluated configuration
claude-sonnet-4-5-20250929-high-32k
Evaluation type
human preference

GPT-5

1,415.44 rating

Confidence interval: 1,409.71 – 1,421.17

15057 votes

Category
hard_prompts
Evaluated configuration
gpt-5-high
Evaluation type
human preference
Arena text preference / hard_prompts_englishhuman preference · 2026-10-15 · rating

Claude Sonnet 4.5

1,471.45 rating

Confidence interval: 1,466.68 – 1,476.22

21147 votes

Category
hard_prompts_english
Evaluated configuration
claude-sonnet-4-5-20250929-high-32k
Evaluation type
human preference

1,470.09 rating

Confidence interval: 1,465.24 – 1,474.95

21308 votes

Category
hard_prompts_english
Evaluated configuration
claude-sonnet-4-5-20250929
Evaluation type
human preference

GPT-5

1,419.40 rating

Confidence interval: 1,411.70 – 1,427.11

6658 votes

Category
hard_prompts_english
Evaluated configuration
gpt-5-high
Evaluation type
human preference
Arena text preference / industry_business_and_management_and_financial_operationshuman preference · 2026-10-15 · rating

Claude Sonnet 4.5

1,440.09 rating

Confidence interval: 1,434.74 – 1,445.43

15836 votes

Category
industry_business_and_management_and_financial_operations
Evaluated configuration
claude-sonnet-4-5-20250929
Evaluation type
human preference

1,432.47 rating

Confidence interval: 1,427.18 – 1,437.77

16036 votes

Category
industry_business_and_management_and_financial_operations
Evaluated configuration
claude-sonnet-4-5-20250929-high-32k
Evaluation type
human preference

GPT-5

1,377.29 rating

Confidence interval: 1,368.91 – 1,385.68

5535 votes

Category
industry_business_and_management_and_financial_operations
Evaluated configuration
gpt-5-high
Evaluation type
human preference
Arena text preference / industry_entertainment_and_sports_and_mediahuman preference · 2026-10-15 · rating

Claude Sonnet 4.5

1,428.39 rating

Confidence interval: 1,422.95 – 1,433.83

15781 votes

Category
industry_entertainment_and_sports_and_media
Evaluated configuration
claude-sonnet-4-5-20250929
Evaluation type
human preference

1,419.87 rating

Confidence interval: 1,414.59 – 1,425.15

16173 votes

Category
industry_entertainment_and_sports_and_media
Evaluated configuration
claude-sonnet-4-5-20250929-high-32k
Evaluation type
human preference

GPT-5

1,378.89 rating

Confidence interval: 1,370.80 – 1,386.97

5876 votes

Category
industry_entertainment_and_sports_and_media
Evaluated configuration
gpt-5-high
Evaluation type
human preference
Arena text preference / industry_legal_and_governmenthuman preference · 2026-10-15 · rating

Claude Sonnet 4.5

1,450.54 rating

Confidence interval: 1,442.56 – 1,458.53

5987 votes

Category
industry_legal_and_government
Evaluated configuration
claude-sonnet-4-5-20250929
Evaluation type
human preference

1,441.42 rating

Confidence interval: 1,433.37 – 1,449.47

5876 votes

Category
industry_legal_and_government
Evaluated configuration
claude-sonnet-4-5-20250929-high-32k
Evaluation type
human preference

GPT-5

1,423.06 rating

Confidence interval: 1,409.02 – 1,437.10

1830 votes

Category
industry_legal_and_government
Evaluated configuration
gpt-5-high
Evaluation type
human preference
Arena text preference / industry_life_and_physical_and_social_sciencehuman preference · 2026-10-15 · rating

Claude Sonnet 4.5

1,447.69 rating

Confidence interval: 1,442.06 – 1,453.32

13610 votes

Category
industry_life_and_physical_and_social_science
Evaluated configuration
claude-sonnet-4-5-20250929
Evaluation type
human preference

1,441.09 rating

Confidence interval: 1,435.54 – 1,446.64

13626 votes

Category
industry_life_and_physical_and_social_science
Evaluated configuration
claude-sonnet-4-5-20250929-high-32k
Evaluation type
human preference

GPT-5

1,402.63 rating

Confidence interval: 1,394.01 – 1,411.24

5133 votes

Category
industry_life_and_physical_and_social_science
Evaluated configuration
gpt-5-high
Evaluation type
human preference
Arena text preference / industry_mathematicalhuman preference · 2026-10-15 · rating

Claude Sonnet 4.5

1,462.46 rating

Confidence interval: 1,453.18 – 1,471.75

4203 votes

Category
industry_mathematical
Evaluated configuration
claude-sonnet-4-5-20250929-high-32k
Evaluation type
human preference

1,446.35 rating

Confidence interval: 1,437.07 – 1,455.62

4316 votes

Category
industry_mathematical
Evaluated configuration
claude-sonnet-4-5-20250929
Evaluation type
human preference

GPT-5

1,404.55 rating

Confidence interval: 1,390.38 – 1,418.73

1767 votes

Category
industry_mathematical
Evaluated configuration
gpt-5-high
Evaluation type
human preference
Arena text preference / industry_medicine_and_healthcarehuman preference · 2026-10-15 · rating

Claude Sonnet 4.5

1,443.62 rating

Confidence interval: 1,435.05 – 1,452.18

5391 votes

Category
industry_medicine_and_healthcare
Evaluated configuration
claude-sonnet-4-5-20250929
Evaluation type
human preference

1,430.00 rating

Confidence interval: 1,421.49 – 1,438.50

5389 votes

Category
industry_medicine_and_healthcare
Evaluated configuration
claude-sonnet-4-5-20250929-high-32k
Evaluation type
human preference

GPT-5

1,413.08 rating

Confidence interval: 1,398.53 – 1,427.63

1783 votes

Category
industry_medicine_and_healthcare
Evaluated configuration
gpt-5-high
Evaluation type
human preference
Arena text preference / industry_software_and_it_serviceshuman preference · 2026-10-15 · rating

Claude Sonnet 4.5

1,472.45 rating

Confidence interval: 1,468.24 – 1,476.66

30336 votes

Category
industry_software_and_it_services
Evaluated configuration
claude-sonnet-4-5-20250929
Evaluation type
human preference

1,470.92 rating

Confidence interval: 1,466.80 – 1,475.04

30441 votes

Category
industry_software_and_it_services
Evaluated configuration
claude-sonnet-4-5-20250929-high-32k
Evaluation type
human preference

GPT-5

1,420.13 rating

Confidence interval: 1,413.72 – 1,426.54

10822 votes

Category
industry_software_and_it_services
Evaluated configuration
gpt-5-high
Evaluation type
human preference
Arena text preference / industry_writing_and_literature_and_languagehuman preference · 2026-10-15 · rating

Claude Sonnet 4.5

1,441.84 rating

Confidence interval: 1,436.84 – 1,446.84

19003 votes

Category
industry_writing_and_literature_and_language
Evaluated configuration
claude-sonnet-4-5-20250929
Evaluation type
human preference

1,431.82 rating

Confidence interval: 1,426.92 – 1,436.71

19381 votes

Category
industry_writing_and_literature_and_language
Evaluated configuration
claude-sonnet-4-5-20250929-high-32k
Evaluation type
human preference

GPT-5

1,378.65 rating

Confidence interval: 1,371.16 – 1,386.14

7039 votes

Category
industry_writing_and_literature_and_language
Evaluated configuration
gpt-5-high
Evaluation type
human preference
Arena text preference / instruction_followinghuman preference · 2026-10-15 · rating

Claude Sonnet 4.5

1,458.94 rating

Confidence interval: 1,454.35 – 1,463.52

24371 votes

Category
instruction_following
Evaluated configuration
claude-sonnet-4-5-20250929
Evaluation type
human preference

1,458.13 rating

Confidence interval: 1,453.64 – 1,462.62

24665 votes

Category
instruction_following
Evaluated configuration
claude-sonnet-4-5-20250929-high-32k
Evaluation type
human preference

GPT-5

1,388.01 rating

Confidence interval: 1,380.95 – 1,395.07

8373 votes

Category
instruction_following
Evaluated configuration
gpt-5-high
Evaluation type
human preference
Arena text preference / japanesehuman preference · 2026-10-15 · rating

Claude Sonnet 4.5

1,389.70 rating

Confidence interval: 1,368.04 – 1,411.37

785 votes

Category
japanese
Evaluated configuration
claude-sonnet-4-5-20250929
Evaluation type
human preference

1,372.66 rating

Confidence interval: 1,350.87 – 1,394.45

779 votes

Category
japanese
Evaluated configuration
claude-sonnet-4-5-20250929-high-32k
Evaluation type
human preference

GPT-5

1,401.60 rating

Confidence interval: 1,380.31 – 1,422.89

914 votes

Category
japanese
Evaluated configuration
gpt-5-high
Evaluation type
human preference
Arena text preference / koreanhuman preference · 2026-10-15 · rating

Claude Sonnet 4.5

1,401.76 rating

Confidence interval: 1,384.78 – 1,418.75

1329 votes

Category
korean
Evaluated configuration
claude-sonnet-4-5-20250929
Evaluation type
human preference

1,366.92 rating

Confidence interval: 1,349.96 – 1,383.88

1355 votes

Category
korean
Evaluated configuration
claude-sonnet-4-5-20250929-high-32k
Evaluation type
human preference

GPT-5

1,357.37 rating

Confidence interval: 1,335.81 – 1,378.93

850 votes

Category
korean
Evaluated configuration
gpt-5-high
Evaluation type
human preference
Arena text preference / longer_queryhuman preference · 2026-10-15 · rating

Claude Sonnet 4.5

1,475.32 rating

Confidence interval: 1,470.73 – 1,479.92

26489 votes

Category
longer_query
Evaluated configuration
claude-sonnet-4-5-20250929
Evaluation type
human preference

1,475.13 rating

Confidence interval: 1,470.59 – 1,479.68

26080 votes

Category
longer_query
Evaluated configuration
claude-sonnet-4-5-20250929-high-32k
Evaluation type
human preference

GPT-5

1,387.13 rating

Confidence interval: 1,379.58 – 1,394.68

6999 votes

Category
longer_query
Evaluated configuration
gpt-5-high
Evaluation type
human preference
Arena text preference / mathhuman preference · 2026-10-15 · rating

Claude Sonnet 4.5

1,448.57 rating

Confidence interval: 1,440.17 – 1,456.96

5072 votes

Category
math
Evaluated configuration
claude-sonnet-4-5-20250929-high-32k
Evaluation type
human preference

1,422.04 rating

Confidence interval: 1,413.68 – 1,430.40

5125 votes

Category
math
Evaluated configuration
claude-sonnet-4-5-20250929
Evaluation type
human preference

GPT-5

1,398.46 rating

Confidence interval: 1,384.79 – 1,412.13

1870 votes

Category
math
Evaluated configuration
gpt-5-high
Evaluation type
human preference
Arena text preference / multi_turnhuman preference · 2026-10-15 · rating

Claude Sonnet 4.5

1,464.93 rating

Confidence interval: 1,459.32 – 1,470.55

14479 votes

Category
multi_turn
Evaluated configuration
claude-sonnet-4-5-20250929
Evaluation type
human preference

1,453.57 rating

Confidence interval: 1,448.02 – 1,459.13

14308 votes

Category
multi_turn
Evaluated configuration
claude-sonnet-4-5-20250929-high-32k
Evaluation type
human preference

GPT-5

1,399.04 rating

Confidence interval: 1,390.24 – 1,407.84

4963 votes

Category
multi_turn
Evaluated configuration
gpt-5-high
Evaluation type
human preference
Arena text preference / non_englishhuman preference · 2026-10-15 · rating

Claude Sonnet 4.5

1,424.92 rating

Confidence interval: 1,421.38 – 1,428.46

46922 votes

Category
non_english
Evaluated configuration
claude-sonnet-4-5-20250929
Evaluation type
human preference

1,417.92 rating

Confidence interval: 1,414.47 – 1,421.36

47676 votes

Category
non_english
Evaluated configuration
claude-sonnet-4-5-20250929-high-32k
Evaluation type
human preference

GPT-5

1,396.93 rating

Confidence interval: 1,391.51 – 1,402.35

18083 votes

Category
non_english
Evaluated configuration
gpt-5-high
Evaluation type
human preference
Arena text preference / overallhuman preference · 2026-10-15 · rating

Claude Sonnet 4.5

1,438.42 rating

Confidence interval: 1,435.57 – 1,441.28

82688 votes

Category
overall
Evaluated configuration
claude-sonnet-4-5-20250929
Evaluation type
human preference

1,434.58 rating

Confidence interval: 1,431.82 – 1,437.34

83940 votes

Category
overall
Evaluated configuration
claude-sonnet-4-5-20250929-high-32k
Evaluation type
human preference

GPT-5

1,406.18 rating

Confidence interval: 1,401.76 – 1,410.60

31722 votes

Category
overall
Evaluated configuration
gpt-5-high
Evaluation type
human preference
Arena text preference / polishhuman preference · 2026-10-15 · rating

Claude Sonnet 4.5

1,422.70 rating

Confidence interval: 1,410.05 – 1,435.34

2225 votes

Category
polish
Evaluated configuration
claude-sonnet-4-5-20250929-high-32k
Evaluation type
human preference

1,419.43 rating

Confidence interval: 1,407.15 – 1,431.72

2316 votes

Category
polish
Evaluated configuration
claude-sonnet-4-5-20250929
Evaluation type
human preference

GPT-5

1,411.55 rating

Confidence interval: 1,397.20 – 1,425.89

1919 votes

Category
polish
Evaluated configuration
gpt-5-high
Evaluation type
human preference
Arena text preference / russianhuman preference · 2026-10-15 · rating

Claude Sonnet 4.5

1,436.35 rating

Confidence interval: 1,429.89 – 1,442.80

9580 votes

Category
russian
Evaluated configuration
claude-sonnet-4-5-20250929
Evaluation type
human preference

1,428.46 rating

Confidence interval: 1,422.06 – 1,434.86

9525 votes

Category
russian
Evaluated configuration
claude-sonnet-4-5-20250929-high-32k
Evaluation type
human preference

GPT-5

1,401.86 rating

Confidence interval: 1,389.28 – 1,414.45

2164 votes

Category
russian
Evaluated configuration
gpt-5-high
Evaluation type
human preference
Arena text preference / spanishhuman preference · 2026-10-15 · rating

Claude Sonnet 4.5

1,457.60 rating

Confidence interval: 1,445.20 – 1,470.00

2499 votes

Category
spanish
Evaluated configuration
claude-sonnet-4-5-20250929
Evaluation type
human preference

1,443.44 rating

Confidence interval: 1,430.99 – 1,455.90

2579 votes

Category
spanish
Evaluated configuration
claude-sonnet-4-5-20250929-high-32k
Evaluation type
human preference

GPT-5

1,387.56 rating

Confidence interval: 1,367.52 – 1,407.60

955 votes

Category
spanish
Evaluated configuration
gpt-5-high
Evaluation type
human preference
Arena text preference / chinesehuman preference · 2026-10-28 · rating

Claude Sonnet 4.5

Not reported for this test

GPT-5

1,421.66 rating

Confidence interval: 1,408.77 – 1,434.55

2245 votes

Category
chinese
Evaluated configuration
GPT-5 (High)
Evaluation type
human preference
Arena text preference / codinghuman preference · 2026-10-28 · rating

Claude Sonnet 4.5

Not reported for this test

GPT-5

1,435.85 rating

Confidence interval: 1,427.95 – 1,443.75

6395 votes

Category
coding
Evaluated configuration
GPT-5 (High)
Evaluation type
human preference
Arena text preference / creative_writinghuman preference · 2026-10-28 · rating

Claude Sonnet 4.5

Not reported for this test

GPT-5

1,363.94 rating

Confidence interval: 1,354.45 – 1,373.43

4242 votes

Category
creative_writing
Evaluated configuration
GPT-5 (High)
Evaluation type
human preference
Arena text preference / englishhuman preference · 2026-10-28 · rating

Claude Sonnet 4.5

Not reported for this test

GPT-5

1,408.95 rating

Confidence interval: 1,403.11 – 1,414.80

13601 votes

Category
english
Evaluated configuration
GPT-5 (High)
Evaluation type
human preference
Arena text preference / exclude_tieshuman preference · 2026-10-28 · rating

Claude Sonnet 4.5

Not reported for this test

GPT-5

1,386.94 rating

Confidence interval: 1,381.05 – 1,392.83

23422 votes

Category
exclude_ties
Evaluated configuration
GPT-5 (High)
Evaluation type
human preference
Arena text preference / experthuman preference · 2026-10-28 · rating

Claude Sonnet 4.5

Not reported for this test

GPT-5

1,418.98 rating

Confidence interval: 1,404.17 – 1,433.79

1695 votes

Category
expert
Evaluated configuration
GPT-5 (High)
Evaluation type
human preference
Arena text preference / frenchhuman preference · 2026-10-28 · rating

Claude Sonnet 4.5

Not reported for this test

GPT-5

1,408.06 rating

Confidence interval: 1,379.56 – 1,436.56

480 votes

Category
french
Evaluated configuration
GPT-5 (High)
Evaluation type
human preference
Arena text preference / germanhuman preference · 2026-10-28 · rating

Claude Sonnet 4.5

Not reported for this test

GPT-5

1,415.92 rating

Confidence interval: 1,393.68 – 1,438.16

747 votes

Category
german
Evaluated configuration
GPT-5 (High)
Evaluation type
human preference
Arena text preference / hard_promptshuman preference · 2026-10-28 · rating

Claude Sonnet 4.5

Not reported for this test

GPT-5

1,415.53 rating

Confidence interval: 1,409.80 – 1,421.26

15058 votes

Category
hard_prompts
Evaluated configuration
GPT-5 (High)
Evaluation type
human preference
Arena text preference / hard_prompts_englishhuman preference · 2026-10-28 · rating

Claude Sonnet 4.5

Not reported for this test

GPT-5

1,419.36 rating

Confidence interval: 1,411.66 – 1,427.07

6656 votes

Category
hard_prompts_english
Evaluated configuration
GPT-5 (High)
Evaluation type
human preference
Arena text preference / industry_business_and_management_and_financial_operationshuman preference · 2026-10-28 · rating

Claude Sonnet 4.5

Not reported for this test

GPT-5

1,377.29 rating

Confidence interval: 1,368.90 – 1,385.67

5536 votes

Category
industry_business_and_management_and_financial_operations
Evaluated configuration
GPT-5 (High)
Evaluation type
human preference
Arena text preference / industry_entertainment_and_sports_and_mediahuman preference · 2026-10-28 · rating

Claude Sonnet 4.5

Not reported for this test

GPT-5

1,378.97 rating

Confidence interval: 1,370.89 – 1,387.06

5879 votes

Category
industry_entertainment_and_sports_and_media
Evaluated configuration
GPT-5 (High)
Evaluation type
human preference
Arena text preference / industry_legal_and_governmenthuman preference · 2026-10-28 · rating

Claude Sonnet 4.5

Not reported for this test

GPT-5

1,423.11 rating

Confidence interval: 1,409.06 – 1,437.15

1829 votes

Category
industry_legal_and_government
Evaluated configuration
GPT-5 (High)
Evaluation type
human preference
Arena text preference / industry_life_and_physical_and_social_sciencehuman preference · 2026-10-28 · rating

Claude Sonnet 4.5

Not reported for this test

GPT-5

1,402.79 rating

Confidence interval: 1,394.17 – 1,411.41

5134 votes

Category
industry_life_and_physical_and_social_science
Evaluated configuration
GPT-5 (High)
Evaluation type
human preference
Arena text preference / industry_mathematicalhuman preference · 2026-10-28 · rating

Claude Sonnet 4.5

Not reported for this test

GPT-5

1,404.52 rating

Confidence interval: 1,390.35 – 1,418.70

1767 votes

Category
industry_mathematical
Evaluated configuration
GPT-5 (High)
Evaluation type
human preference
Arena text preference / industry_medicine_and_healthcarehuman preference · 2026-10-28 · rating

Claude Sonnet 4.5

Not reported for this test

GPT-5

1,413.15 rating

Confidence interval: 1,398.60 – 1,427.70

1784 votes

Category
industry_medicine_and_healthcare
Evaluated configuration
GPT-5 (High)
Evaluation type
human preference
Arena text preference / industry_software_and_it_serviceshuman preference · 2026-10-28 · rating

Claude Sonnet 4.5

Not reported for this test

GPT-5

1,419.93 rating

Confidence interval: 1,413.52 – 1,426.34

10818 votes

Category
industry_software_and_it_services
Evaluated configuration
GPT-5 (High)
Evaluation type
human preference
Arena text preference / industry_writing_and_literature_and_languagehuman preference · 2026-10-28 · rating

Claude Sonnet 4.5

Not reported for this test

GPT-5

1,378.52 rating

Confidence interval: 1,371.03 – 1,386.01

7041 votes

Category
industry_writing_and_literature_and_language
Evaluated configuration
GPT-5 (High)
Evaluation type
human preference
Arena text preference / instruction_followinghuman preference · 2026-10-28 · rating

Claude Sonnet 4.5

Not reported for this test

GPT-5

1,388.30 rating

Confidence interval: 1,381.24 – 1,395.36

8376 votes

Category
instruction_following
Evaluated configuration
GPT-5 (High)
Evaluation type
human preference
Arena text preference / japanesehuman preference · 2026-10-28 · rating

Claude Sonnet 4.5

Not reported for this test

GPT-5

1,402.19 rating

Confidence interval: 1,380.88 – 1,423.51

915 votes

Category
japanese
Evaluated configuration
GPT-5 (High)
Evaluation type
human preference
Arena text preference / koreanhuman preference · 2026-10-28 · rating

Claude Sonnet 4.5

Not reported for this test

GPT-5

1,358.63 rating

Confidence interval: 1,337.10 – 1,380.15

852 votes

Category
korean
Evaluated configuration
GPT-5 (High)
Evaluation type
human preference
Arena text preference / longer_queryhuman preference · 2026-10-28 · rating

Claude Sonnet 4.5

Not reported for this test

GPT-5

1,387.20 rating

Confidence interval: 1,379.65 – 1,394.74

6999 votes

Category
longer_query
Evaluated configuration
GPT-5 (High)
Evaluation type
human preference
Arena text preference / mathhuman preference · 2026-10-28 · rating

Claude Sonnet 4.5

Not reported for this test

GPT-5

1,398.25 rating

Confidence interval: 1,384.57 – 1,411.93

1868 votes

Category
math
Evaluated configuration
GPT-5 (High)
Evaluation type
human preference
Arena text preference / multi_turnhuman preference · 2026-10-28 · rating

Claude Sonnet 4.5

Not reported for this test

GPT-5

1,399.09 rating

Confidence interval: 1,390.29 – 1,407.89

4961 votes

Category
multi_turn
Evaluated configuration
GPT-5 (High)
Evaluation type
human preference
Arena text preference / non_englishhuman preference · 2026-10-28 · rating

Claude Sonnet 4.5

Not reported for this test

GPT-5

1,396.97 rating

Confidence interval: 1,391.55 – 1,402.40

18084 votes

Category
non_english
Evaluated configuration
GPT-5 (High)
Evaluation type
human preference
Arena text preference / overallhuman preference · 2026-10-28 · rating

Claude Sonnet 4.5

Not reported for this test

GPT-5

1,406.15 rating

Confidence interval: 1,401.73 – 1,410.57

31718 votes

Category
overall
Evaluated configuration
GPT-5 (High)
Evaluation type
human preference
Arena text preference / polishhuman preference · 2026-10-28 · rating

Claude Sonnet 4.5

Not reported for this test

GPT-5

1,411.50 rating

Confidence interval: 1,397.16 – 1,425.85

1919 votes

Category
polish
Evaluated configuration
GPT-5 (High)
Evaluation type
human preference
Arena text preference / russianhuman preference · 2026-10-28 · rating

Claude Sonnet 4.5

Not reported for this test

GPT-5

1,402.48 rating

Confidence interval: 1,389.90 – 1,415.06

2164 votes

Category
russian
Evaluated configuration
GPT-5 (High)
Evaluation type
human preference
Arena text preference / spanishhuman preference · 2026-10-28 · rating

Claude Sonnet 4.5

Not reported for this test

GPT-5

1,388.01 rating

Confidence interval: 1,367.98 – 1,408.03

956 votes

Category
spanish
Evaluated configuration
GPT-5 (High)
Evaluation type
human preference

* Provisional data