geekradar
Int
Explore GeekRadar

More to explore

Good gear All topics About us Say hello
GeekRadar / The comparison desk
COMPARE.
AI models
anthropic

Claude Opus 4.6

AI model Released

API mode
Chat*
Context window
1,000,000 tokens
Maximum output
128,000 tokens
Open weights
No

Capability & preference

General capability · ECI

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

Claude Opus 4.6 155.9* Claude Opus 4.6 · 95% interval 154.1–158.0

Human preference · Arena

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

Claude Opus 4.6 1496.2 Claude Opus 4.6 · 95% interval 1493.1–1499.4 · 89,324 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 Opus 4.6

Radar task score

56.6* / 100

4/4 core tests · experimental v1

8 benchmark results for this configuration

Scientific reasoning

GPQA Diamond · 25%

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

Claude Opus 4.6 88.4*% Approx. 95% interval: 83.9–92.9%

Novel problem solving

Mystery Game Puzzles · 25%

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

Claude Opus 4.6 25.0*% Approx. 95% interval: 16.5–33.5%

Advanced mathematics

FrontierMath Tiers 1-3 · 25%

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

Claude Opus 4.6 66.0*% Approx. 95% interval: 60.5–71.5%

Factual answers

SimpleQA Verified · 25%

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

Claude Opus 4.6 47.0*% Approx. 95% interval: 43.9–50.1%

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 Opus 4.6 33.3*% Approx. 95% interval: 13.7–53.0%

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 Opus 4.6 91.1*% Approx. 95% interval: 82.7–99.5%

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 Opus 4.6 26.8*% Approx. 95% interval: 13.1–40.6%

Visual spatial reasoning

Furniture Assembly · Separate result

Assembly-error detection from photos and manuals, with zoom and Python tools. Uses model-assisted grading; not a general vision score.

Claude Opus 4.6 28.3*% Approx. 95% interval: 16.8–39.8%

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.

Benchmark results

  • Agentic coding SWE-bench Pro · version unreported · reported score · Not specified · Not specified · max · Previously reported result; re-evaluation not stated53.40*Platform unreported Compare results
    Test conditions
    Metric
    reported score*
    Checkpoint version
    Not specified*
    Reasoning setting
    max*
    Statistic
    Previously reported result; re-evaluation not stated*
    Task set
    Not specified*
    Harness
    Not specified*
  • Agentic terminal coding Terminal Bench 2.1 (Terminus) · 2.1 · reported score · Terminus · max78.20*Platform unreported Compare results
    Test conditions
    Metric
    reported score*
    Checkpoint version
    Not specified*
    Reasoning setting
    max*
    Harness
    Terminus*
  • Competitive coding LiveCodeBench v6 · 6 · reported score · max88.80*Platform unreported Compare results
    Test conditions
    Metric
    reported score*
    Checkpoint version
    Not specified*
    Reasoning setting
    max*
  • Computer use OSWorld-Verified · version unreported · reported score · max72.70*Platform unreported Compare results
    Test conditions
    Metric
    reported score*
    Checkpoint version
    Not specified*
    Reasoning setting
    max*
  • Document intelligence OmniDocBench 1.5 · 1.5 · reported score · max86.60*Platform unreported Compare results
    Test conditions
    Metric
    reported score*
    Checkpoint version
    Not specified*
    Reasoning setting
    max*
  • Embodied intelligence ERQA · version unreported · reported score · max40.80*Platform unreported Compare results
    Test conditions
    Metric
    reported score*
    Checkpoint version
    Not specified*
    Reasoning setting
    max*
  • General visual reasoning BabyVision · version unreported · reported score · Code interpreter disabled · max12.60*Platform unreported Compare results
    Test conditions
    Metric
    reported score*
    Checkpoint version
    Not specified*
    Tool access
    Code interpreter disabled*
    Reasoning setting
    max*
  • Instruction following IFBench · version unreported · reported score · max62.50*Platform unreported Compare results
    Test conditions
    Metric
    reported score*
    Checkpoint version
    Not specified*
    Reasoning setting
    max*
  • Long-horizon office work CoWorkBench · version unreported · reported score · Internal · max68.20*Platform unreported Compare results
    Test conditions
    Metric
    reported score*
    Checkpoint version
    Not specified*
    Reasoning setting
    max*
    Task set
    Internal*
  • Mobile use AndroidWorld · version unreported · reported score · max62*Platform unreported Compare results
    Test conditions
    Metric
    reported score*
    Checkpoint version
    Not specified*
    Reasoning setting
    max*
  • Multidisciplinary reasoning HLE · version unreported · reported score · max40*Platform unreported Compare results
    Test conditions
    Metric
    reported score*
    Checkpoint version
    Not specified*
    Reasoning setting
    max*
    Judge
    GPT-4o*
  • Multimodal software engineering SWE-MM · version unreported · reported score · Claude Code · Public SWE-bench Multimodal dev split with Opus 4.7 system-card appendix 8.3 modifications · max27.10*Platform unreported Compare results
    Test conditions
    Metric
    reported score*
    Checkpoint version
    Not specified*
    Reasoning setting
    max*
    Harness
    Claude Code*
    Task set
    Public SWE-bench Multimodal dev split with Opus 4.7 system-card appendix 8.3 modifications*
  • Multimodal tool use ClawEval-MM · version unreported · average score · max · Mean benchmark score over three trials54.70*Platform unreported Compare results
    Test conditions
    Metric
    average score*
    Checkpoint version
    Not specified*
    Reasoning setting
    max*
    Statistic
    Mean benchmark score over three trials*
  • Multimodal tool use ClawEval-MM · version unreported · pass@3 · max · At least one success in three trials52.50* %Platform unreported Compare results
    Test conditions
    Metric
    pass@3*
    Checkpoint version
    Not specified*
    Reasoning setting
    max*
    Statistic
    At least one success in three trials*
  • Real-world perception RealWorldQA · version unreported · reported score · max73.90*Platform unreported Compare results
    Test conditions
    Metric
    reported score*
    Checkpoint version
    Not specified*
    Reasoning setting
    max*
  • Repo-level code generation NL2Repo-Bench · version unreported · reported score · Claude Code · Repository acquisition commands disabled · max47.60*Platform unreported Compare results
    Test conditions
    Metric
    reported score*
    Checkpoint version
    Not specified*
    Reasoning setting
    max*
    Harness
    Claude Code*
    Tool access
    Repository acquisition commands disabled*
  • Scientific chart analysis CharXiv (RQ) · version unreported · reported score · Code interpreter disabled · max66*Platform unreported Compare results
    Test conditions
    Metric
    reported score*
    Checkpoint version
    Not specified*
    Tool access
    Code interpreter disabled*
    Reasoning setting
    max*
    Annotation set
    Corrected ground-truth annotations*
  • Scientific reasoning GPQA Diamond · version unreported · reported score · max91.30*Platform unreported Compare results
    Test conditions
    Metric
    reported score*
    Checkpoint version
    Not specified*
    Reasoning setting
    max*
  • Software engineering QwenSWEBench · version unreported · reported score · Claude Code · Internal · max · avg@363.80*Platform unreported Compare results
    Test conditions
    Metric
    reported score*
    Checkpoint version
    Not specified*
    Reasoning setting
    max*
    Harness
    Claude Code*
    Statistic
    avg@3*
    Timeout
    8 hours*
    Task set
    Internal*
    Maximum output
    32,768*
    Temperature
    1.0*
    Context length
    256K tokens*
  • Visual math problem solving MathVision · version unreported · reported score · Code interpreter disabled · max65.50*Platform unreported Compare results
    Test conditions
    Metric
    reported score*
    Checkpoint version
    Not specified*
    Tool access
    Code interpreter disabled*
    Reasoning setting
    max*
    Annotation set
    Corrected ground-truth annotations*
    Prompt
    Best of two prompts, with and without boxed final-answer formatting*

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 · 2026-10-28 · rating

Claude Opus 4.6

1,550.64 rating

Confidence interval: 1,542.06 – 1,559.21

5913 votes

Category
chinese
Evaluated configuration
Claude Opus 4.6 (High)
Evaluation type
human preference

1,546.18 rating

Confidence interval: 1,537.76 – 1,554.59

6141 votes

Category
chinese
Evaluated configuration
Claude Opus 4.6
Evaluation type
human preference
Arena text preference / codinghuman preference · 2026-10-28 · rating

Claude Opus 4.6

1,533.43 rating

Confidence interval: 1,528.39 – 1,538.48

24611 votes

Category
coding
Evaluated configuration
Claude Opus 4.6
Evaluation type
human preference

1,531.35 rating

Confidence interval: 1,526.10 – 1,536.60

21789 votes

Category
coding
Evaluated configuration
Claude Opus 4.6 (High)
Evaluation type
human preference
Arena text preference / creative_writinghuman preference · 2026-10-28 · rating

Claude Opus 4.6

1,504.50 rating

Confidence interval: 1,498.41 – 1,510.59

15580 votes

Category
creative_writing
Evaluated configuration
Claude Opus 4.6 (High)
Evaluation type
human preference

1,483.07 rating

Confidence interval: 1,477.14 – 1,489.00

15812 votes

Category
creative_writing
Evaluated configuration
Claude Opus 4.6
Evaluation type
human preference
Arena text preference / englishhuman preference · 2026-10-28 · rating

Claude Opus 4.6

1,511.58 rating

Confidence interval: 1,507.20 – 1,515.95

37010 votes

Category
english
Evaluated configuration
Claude Opus 4.6 (High)
Evaluation type
human preference

1,505.74 rating

Confidence interval: 1,501.48 – 1,509.99

39613 votes

Category
english
Evaluated configuration
Claude Opus 4.6
Evaluation type
human preference
Arena text preference / exclude_tieshuman preference · 2026-10-28 · rating

Claude Opus 4.6

1,516.99 rating

Confidence interval: 1,512.66 – 1,521.32

64400 votes

Category
exclude_ties
Evaluated configuration
Claude Opus 4.6 (High)
Evaluation type
human preference

1,511.50 rating

Confidence interval: 1,507.26 – 1,515.75

67789 votes

Category
exclude_ties
Evaluated configuration
Claude Opus 4.6
Evaluation type
human preference
Arena text preference / experthuman preference · 2026-10-28 · rating

Claude Opus 4.6

1,542.82 rating

Confidence interval: 1,535.18 – 1,550.45

7698 votes

Category
expert
Evaluated configuration
Claude Opus 4.6 (High)
Evaluation type
human preference

1,536.89 rating

Confidence interval: 1,529.71 – 1,544.08

9181 votes

Category
expert
Evaluated configuration
Claude Opus 4.6
Evaluation type
human preference
Arena text preference / frenchhuman preference · 2026-10-28 · rating

Claude Opus 4.6

1,512.10 rating

Confidence interval: 1,499.06 – 1,525.14

2770 votes

Category
french
Evaluated configuration
Claude Opus 4.6 (High)
Evaluation type
human preference

1,498.46 rating

Confidence interval: 1,485.76 – 1,511.16

2971 votes

Category
french
Evaluated configuration
Claude Opus 4.6
Evaluation type
human preference
Arena text preference / germanhuman preference · 2026-10-28 · rating

Claude Opus 4.6

1,501.83 rating

Confidence interval: 1,484.70 – 1,518.97

1336 votes

Category
german
Evaluated configuration
Claude Opus 4.6 (High)
Evaluation type
human preference

1,490.18 rating

Confidence interval: 1,473.37 – 1,506.99

1397 votes

Category
german
Evaluated configuration
Claude Opus 4.6
Evaluation type
human preference
Arena text preference / hard_promptshuman preference · 2026-10-28 · rating

Claude Opus 4.6

1,524.65 rating

Confidence interval: 1,520.69 – 1,528.61

53763 votes

Category
hard_prompts
Evaluated configuration
Claude Opus 4.6 (High)
Evaluation type
human preference

1,521.67 rating

Confidence interval: 1,517.82 – 1,525.52

57870 votes

Category
hard_prompts
Evaluated configuration
Claude Opus 4.6
Evaluation type
human preference
Arena text preference / hard_prompts_englishhuman preference · 2026-10-28 · rating

Claude Opus 4.6

1,530.45 rating

Confidence interval: 1,525.30 – 1,535.60

24033 votes

Category
hard_prompts_english
Evaluated configuration
Claude Opus 4.6 (High)
Evaluation type
human preference

1,526.84 rating

Confidence interval: 1,521.86 – 1,531.81

26221 votes

Category
hard_prompts_english
Evaluated configuration
Claude Opus 4.6
Evaluation type
human preference
Arena text preference / industry_business_and_management_and_financial_operationshuman preference · 2026-10-28 · rating

Claude Opus 4.6

1,497.08 rating

Confidence interval: 1,491.40 – 1,502.75

16500 votes

Category
industry_business_and_management_and_financial_operations
Evaluated configuration
Claude Opus 4.6 (High)
Evaluation type
human preference

1,495.28 rating

Confidence interval: 1,489.73 – 1,500.83

17526 votes

Category
industry_business_and_management_and_financial_operations
Evaluated configuration
Claude Opus 4.6
Evaluation type
human preference
Arena text preference / industry_entertainment_and_sports_and_mediahuman preference · 2026-10-28 · rating

Claude Opus 4.6

1,492.12 rating

Confidence interval: 1,486.52 – 1,497.72

19162 votes

Category
industry_entertainment_and_sports_and_media
Evaluated configuration
Claude Opus 4.6 (High)
Evaluation type
human preference

1,477.97 rating

Confidence interval: 1,472.54 – 1,483.39

20063 votes

Category
industry_entertainment_and_sports_and_media
Evaluated configuration
Claude Opus 4.6
Evaluation type
human preference
Arena text preference / industry_legal_and_governmenthuman preference · 2026-10-28 · rating

Claude Opus 4.6

1,511.36 rating

Confidence interval: 1,503.36 – 1,519.36

6874 votes

Category
industry_legal_and_government
Evaluated configuration
Claude Opus 4.6 (High)
Evaluation type
human preference

1,508.55 rating

Confidence interval: 1,500.67 – 1,516.43

7111 votes

Category
industry_legal_and_government
Evaluated configuration
Claude Opus 4.6
Evaluation type
human preference
Arena text preference / industry_life_and_physical_and_social_sciencehuman preference · 2026-10-28 · rating

Claude Opus 4.6

1,518.45 rating

Confidence interval: 1,512.48 – 1,524.43

14037 votes

Category
industry_life_and_physical_and_social_science
Evaluated configuration
Claude Opus 4.6 (High)
Evaluation type
human preference

1,511.39 rating

Confidence interval: 1,505.54 – 1,517.24

14782 votes

Category
industry_life_and_physical_and_social_science
Evaluated configuration
Claude Opus 4.6
Evaluation type
human preference
Arena text preference / industry_mathematicalhuman preference · 2026-10-28 · rating

Claude Opus 4.6

1,524.95 rating

Confidence interval: 1,515.24 – 1,534.65

4411 votes

Category
industry_mathematical
Evaluated configuration
Claude Opus 4.6 (High)
Evaluation type
human preference

1,515.87 rating

Confidence interval: 1,506.72 – 1,525.03

5044 votes

Category
industry_mathematical
Evaluated configuration
Claude Opus 4.6
Evaluation type
human preference
Arena text preference / industry_medicine_and_healthcarehuman preference · 2026-10-28 · rating

Claude Opus 4.6

1,503.97 rating

Confidence interval: 1,495.66 – 1,512.28

6343 votes

Category
industry_medicine_and_healthcare
Evaluated configuration
Claude Opus 4.6 (High)
Evaluation type
human preference

1,497.28 rating

Confidence interval: 1,489.12 – 1,505.44

6623 votes

Category
industry_medicine_and_healthcare
Evaluated configuration
Claude Opus 4.6
Evaluation type
human preference
Arena text preference / industry_software_and_it_serviceshuman preference · 2026-10-28 · rating

Claude Opus 4.6

1,526.61 rating

Confidence interval: 1,522.14 – 1,531.07

34720 votes

Category
industry_software_and_it_services
Evaluated configuration
Claude Opus 4.6
Evaluation type
human preference

1,525.43 rating

Confidence interval: 1,520.86 – 1,530.01

31981 votes

Category
industry_software_and_it_services
Evaluated configuration
Claude Opus 4.6 (High)
Evaluation type
human preference
Arena text preference / industry_writing_and_literature_and_languagehuman preference · 2026-10-28 · rating

Claude Opus 4.6

1,501.83 rating

Confidence interval: 1,496.58 – 1,507.07

21945 votes

Category
industry_writing_and_literature_and_language
Evaluated configuration
Claude Opus 4.6 (High)
Evaluation type
human preference

1,491.63 rating

Confidence interval: 1,486.48 – 1,496.78

22357 votes

Category
industry_writing_and_literature_and_language
Evaluated configuration
Claude Opus 4.6
Evaluation type
human preference
Arena text preference / instruction_followinghuman preference · 2026-10-28 · rating

Claude Opus 4.6

1,520.85 rating

Confidence interval: 1,515.95 – 1,525.76

27276 votes

Category
instruction_following
Evaluated configuration
Claude Opus 4.6 (High)
Evaluation type
human preference

1,509.20 rating

Confidence interval: 1,504.50 – 1,513.91

29906 votes

Category
instruction_following
Evaluated configuration
Claude Opus 4.6
Evaluation type
human preference
Arena text preference / japanesehuman preference · 2026-10-28 · rating

Claude Opus 4.6

1,487.45 rating

Confidence interval: 1,466.81 – 1,508.08

986 votes

Category
japanese
Evaluated configuration
Claude Opus 4.6 (High)
Evaluation type
human preference

1,473.85 rating

Confidence interval: 1,453.56 – 1,494.13

1027 votes

Category
japanese
Evaluated configuration
Claude Opus 4.6
Evaluation type
human preference
Arena text preference / koreanhuman preference · 2026-10-28 · rating

Claude Opus 4.6

1,463.09 rating

Confidence interval: 1,446.52 – 1,479.65

1575 votes

Category
korean
Evaluated configuration
Claude Opus 4.6
Evaluation type
human preference

1,450.08 rating

Confidence interval: 1,432.96 – 1,467.21

1417 votes

Category
korean
Evaluated configuration
Claude Opus 4.6 (High)
Evaluation type
human preference
Arena text preference / longer_queryhuman preference · 2026-10-28 · rating

Claude Opus 4.6

1,517.53 rating

Confidence interval: 1,512.86 – 1,522.21

35795 votes

Category
longer_query
Evaluated configuration
Claude Opus 4.6 (High)
Evaluation type
human preference

1,514.00 rating

Confidence interval: 1,509.46 – 1,518.55

38765 votes

Category
longer_query
Evaluated configuration
Claude Opus 4.6
Evaluation type
human preference
Arena text preference / mathhuman preference · 2026-10-28 · rating

Claude Opus 4.6

1,517.54 rating

Confidence interval: 1,508.13 – 1,526.96

4364 votes

Category
math
Evaluated configuration
Claude Opus 4.6 (High)
Evaluation type
human preference

1,509.11 rating

Confidence interval: 1,500.02 – 1,518.20

4841 votes

Category
math
Evaluated configuration
Claude Opus 4.6
Evaluation type
human preference
Arena text preference / multi_turnhuman preference · 2026-10-28 · rating

Claude Opus 4.6

1,512.16 rating

Confidence interval: 1,506.09 – 1,518.24

14319 votes

Category
multi_turn
Evaluated configuration
Claude Opus 4.6 (High)
Evaluation type
human preference

1,505.57 rating

Confidence interval: 1,499.72 – 1,511.43

15566 votes

Category
multi_turn
Evaluated configuration
Claude Opus 4.6
Evaluation type
human preference
Arena text preference / non_englishhuman preference · 2026-10-28 · rating

Claude Opus 4.6

1,487.55 rating

Confidence interval: 1,483.59 – 1,491.50

47491 votes

Category
non_english
Evaluated configuration
Claude Opus 4.6 (High)
Evaluation type
human preference

1,483.86 rating

Confidence interval: 1,479.98 – 1,487.74

49706 votes

Category
non_english
Evaluated configuration
Claude Opus 4.6
Evaluation type
human preference
Arena text preference / overallhuman preference · 2026-10-28 · rating

Claude Opus 4.6

1,501.72 rating

Confidence interval: 1,498.48 – 1,504.97

84512 votes

Category
overall
Evaluated configuration
Claude Opus 4.6 (High)
Evaluation type
human preference

1,496.25 rating

Confidence interval: 1,493.09 – 1,499.41

89324 votes

Category
overall
Evaluated configuration
Claude Opus 4.6
Evaluation type
human preference
Arena text preference / polishhuman preference · 2026-10-28 · rating

Claude Opus 4.6

1,497.96 rating

Confidence interval: 1,482.97 – 1,512.95

1595 votes

Category
polish
Evaluated configuration
Claude Opus 4.6 (High)
Evaluation type
human preference

1,488.54 rating

Confidence interval: 1,473.38 – 1,503.69

1635 votes

Category
polish
Evaluated configuration
Claude Opus 4.6
Evaluation type
human preference
Arena text preference / russianhuman preference · 2026-10-28 · rating

Claude Opus 4.6

1,494.82 rating

Confidence interval: 1,488.18 – 1,501.45

10205 votes

Category
russian
Evaluated configuration
Claude Opus 4.6
Evaluation type
human preference

1,494.59 rating

Confidence interval: 1,487.83 – 1,501.35

9833 votes

Category
russian
Evaluated configuration
Claude Opus 4.6 (High)
Evaluation type
human preference
Arena text preference / spanishhuman preference · 2026-10-28 · rating

Claude Opus 4.6

1,509.00 rating

Confidence interval: 1,496.90 – 1,521.10

2974 votes

Category
spanish
Evaluated configuration
Claude Opus 4.6
Evaluation type
human preference

1,505.47 rating

Confidence interval: 1,493.19 – 1,517.75

2864 votes

Category
spanish
Evaluated configuration
Claude Opus 4.6 (High)
Evaluation type
human preference

Run it locally

Inference · memory estimate

Claude Opus 4.6

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

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 Opus 4.6

Claude Opus 4.6 / Pricing tiers 5
Provider / scopeRateCurrencyAmount / unit
AWS BedrockClaude Opus 4.6Batch Global cross-region InputUSD2.5 / 1M tokens
OutputUSD12.5 / 1M tokens
AWS BedrockClaude Opus 4.6Standard Global cross-region Cached inputUSD0.5 / 1M tokens
InputUSD5 / 1M tokens
OutputUSD25 / 1M tokens
Google Vertex AIClaude Opus 4.6Batch Token tier: 0 – 200,000 Cached inputUSD0.25 / 1M tokens
InputUSD2.5 / 1M tokens
OutputUSD12.5 / 1M tokens
Google Vertex AIClaude Opus 4.6Standard Token tier: 0 – 200,000 Cached inputUSD0.5 / 1M tokens
InputUSD5 / 1M tokens
OutputUSD25 / 1M tokens
Google Vertex AIClaude Opus 4.6Standard Token tier: 200,000 – No stated upper limit Cached inputUSD0.5 / 1M tokens
InputUSD5 / 1M tokens
OutputUSD25 / 1M tokens

Capabilities

Context window
1,000,000 tokens
Maximum input
1,000,000 tokens*
Maximum output
128,000 tokens
Reasoning
Yes
Tool calling
Yes
Structured output
Yes*
Open weights
No
Knowledge cutoff
2025-05-31
API mode
Chat*
Image input
Yes*
PDF input
Yes*
Prompt caching
Yes*
Tool choice
Yes*
Web search
Yes*
Model license
Proprietary

Input modalities

  • text
  • image
  • pdf

Output modalities

  • text

System configuration

Detailed specifications

Specifications 1 · 35 fields
cache creation input token cost
6.25e-6*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-opus-4-6*
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
1.0e-5*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-opus-4-6*
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
3.125e-6*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-opus-4-6*
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
5.0e-7*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-opus-4-6*
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
2.5e-7*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-opus-4-6*
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
5.0e-6*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-opus-4-6*
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
2.5e-6*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-opus-4-6*
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-opus-4-6*
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
1,000,000*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-opus-4-6*
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
anthropic*
mode
chat*
full_native_model_id
claude-opus-4-6*
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
anthropic*
mode
chat*
full_native_model_id
claude-opus-4-6*
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-opus-4-6*
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
2.5e-5*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-opus-4-6*
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
1.25e-5*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-opus-4-6*
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
4,096*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-opus-4-6*
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.*
provider specific entry
us
1.1*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-opus-4-6*
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-opus-4-6*
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 adaptive thinking
Yes*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-opus-4-6*
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 anthropic compaction
Yes*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-opus-4-6*
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
No*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-opus-4-6*
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-opus-4-6*
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-opus-4-6*
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 legacy thinking
Yes*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-opus-4-6*
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 max reasoning effort
Yes*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-opus-4-6*
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-opus-4-6*
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 output config
Yes*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-opus-4-6*
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-opus-4-6*
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-opus-4-6*
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-opus-4-6*
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-opus-4-6*
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 speed
Yes*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-opus-4-6*
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 thinking cache preservation
Yes*
Scope and conditions
native_context
provider
anthropic*
mode
chat*
full_native_model_id
claude-opus-4-6*
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-opus-4-6*
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-opus-4-6*
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
anthropic*
mode
chat*
full_native_model_id
claude-opus-4-6*
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 · 20 fields
Visual math problem solving MathVision
suite
Visual math problem solving MathVision
workload
reported score · Code interpreter disabled · max
score
65.5*
Scope and conditions
Test conditions
Metric
reported score*
Checkpoint version
Not specified*
Tool access
Code interpreter disabled*
Reasoning setting
max*
Annotation set
Corrected ground-truth annotations*
Prompt
Best of two prompts, with and without boxed final-answer formatting*
General visual reasoning BabyVision
suite
General visual reasoning BabyVision
workload
reported score · Code interpreter disabled · max
score
12.6*
Scope and conditions
Test conditions
Metric
reported score*
Checkpoint version
Not specified*
Tool access
Code interpreter disabled*
Reasoning setting
max*
Instruction following IFBench
suite
Instruction following IFBench
workload
reported score · max
score
62.5*
Scope and conditions
Test conditions
Metric
reported score*
Checkpoint version
Not specified*
Reasoning setting
max*
Scientific chart analysis CharXiv (RQ)
suite
Scientific chart analysis CharXiv (RQ)
workload
reported score · Code interpreter disabled · max
score
66.0*
Scope and conditions
Test conditions
Metric
reported score*
Checkpoint version
Not specified*
Tool access
Code interpreter disabled*
Reasoning setting
max*
Annotation set
Corrected ground-truth annotations*
Scientific reasoning GPQA Diamond
suite
Scientific reasoning GPQA Diamond
workload
reported score · max
score
91.3*
Scope and conditions
Test conditions
Metric
reported score*
Checkpoint version
Not specified*
Reasoning setting
max*
Document intelligence OmniDocBench 1.5
suite
Document intelligence OmniDocBench 1.5
version
1.5
workload
reported score · max
score
86.6*
Scope and conditions
Test conditions
Metric
reported score*
Checkpoint version
Not specified*
Reasoning setting
max*
Multidisciplinary reasoning HLE
suite
Multidisciplinary reasoning HLE
workload
reported score · max
score
40.0*
Scope and conditions
Test conditions
Metric
reported score*
Checkpoint version
Not specified*
Reasoning setting
max*
Judge
GPT-4o*
Real-world perception RealWorldQA
suite
Real-world perception RealWorldQA
workload
reported score · max
score
73.9*
Scope and conditions
Test conditions
Metric
reported score*
Checkpoint version
Not specified*
Reasoning setting
max*
Embodied intelligence ERQA
suite
Embodied intelligence ERQA
workload
reported score · max
score
40.8*
Scope and conditions
Test conditions
Metric
reported score*
Checkpoint version
Not specified*
Reasoning setting
max*
Competitive coding LiveCodeBench v6
suite
Competitive coding LiveCodeBench v6
version
6
workload
reported score · max
score
88.8*
Scope and conditions
Test conditions
Metric
reported score*
Checkpoint version
Not specified*
Reasoning setting
max*
Computer use OSWorld-Verified
suite
Computer use OSWorld-Verified
workload
reported score · max
score
72.7*
Scope and conditions
Test conditions
Metric
reported score*
Checkpoint version
Not specified*
Reasoning setting
max*
Agentic terminal coding Terminal Bench 2.1 (Terminus)
suite
Agentic terminal coding Terminal Bench 2.1 (Terminus)
version
2.1
workload
reported score · Terminus · max
score
78.2*
Scope and conditions
Test conditions
Metric
reported score*
Checkpoint version
Not specified*
Reasoning setting
max*
Harness
Terminus*
Agentic coding SWE-bench Pro
suite
Agentic coding SWE-bench Pro
workload
reported score · Not specified · Not specified · max · Previously reported result; re-evaluation not stated
score
53.4*
Scope and conditions
Test conditions
Metric
reported score*
Checkpoint version
Not specified*
Reasoning setting
max*
Statistic
Previously reported result; re-evaluation not stated*
Task set
Not specified*
Harness
Not specified*
Repo-level code generation NL2Repo-Bench
suite
Repo-level code generation NL2Repo-Bench
workload
reported score · Claude Code · Repository acquisition commands disabled · max
score
47.6*
Scope and conditions
Test conditions
Metric
reported score*
Checkpoint version
Not specified*
Reasoning setting
max*
Harness
Claude Code*
Tool access
Repository acquisition commands disabled*
Mobile use AndroidWorld
suite
Mobile use AndroidWorld
workload
reported score · max
score
62.0*
Scope and conditions
Test conditions
Metric
reported score*
Checkpoint version
Not specified*
Reasoning setting
max*
Software engineering QwenSWEBench
suite
Software engineering QwenSWEBench
workload
reported score · Claude Code · Internal · max · avg@3
score
63.8*
Scope and conditions
Test conditions
Metric
reported score*
Checkpoint version
Not specified*
Reasoning setting
max*
Harness
Claude Code*
Statistic
avg@3*
Timeout
8 hours*
Task set
Internal*
Maximum output
32,768*
Temperature
1.0*
Context length
256K tokens*
Multimodal tool use ClawEval-MM
suite
Multimodal tool use ClawEval-MM
workload
pass@3 · max · At least one success in three trials
score
52.5*
unit
%
Scope and conditions
Test conditions
Metric
pass@3*
Checkpoint version
Not specified*
Reasoning setting
max*
Statistic
At least one success in three trials*
Multimodal tool use ClawEval-MM
suite
Multimodal tool use ClawEval-MM
workload
average score · max · Mean benchmark score over three trials
score
54.7*
Scope and conditions
Test conditions
Metric
average score*
Checkpoint version
Not specified*
Reasoning setting
max*
Statistic
Mean benchmark score over three trials*
Multimodal software engineering SWE-MM
suite
Multimodal software engineering SWE-MM
workload
reported score · Claude Code · Public SWE-bench Multimodal dev split with Opus 4.7 system-card appendix 8.3 modifications · max
score
27.1*
Scope and conditions
Test conditions
Metric
reported score*
Checkpoint version
Not specified*
Reasoning setting
max*
Harness
Claude Code*
Task set
Public SWE-bench Multimodal dev split with Opus 4.7 system-card appendix 8.3 modifications*
Long-horizon office work CoWorkBench
suite
Long-horizon office work CoWorkBench
workload
reported score · Internal · max
score
68.2*
Scope and conditions
Test conditions
Metric
reported score*
Checkpoint version
Not specified*
Reasoning setting
max*
Task set
Internal*

* Provisional data