GeekRadar / The comparison desk

COMPARE.

AI models

openai

GPT OSS 20B

AI model Released Aug 5, 2025

Context window: 131,072 tokens
Maximum output: 32,768 tokens
Open weights: Yes

Capability & preference

General capability · ECI

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

GPT OSS 20B 135.0^* gpt-oss-20b · 95% interval 132.4–137.5

Human preference · Arena

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

GPT OSS 20B 1287.3 gpt-oss-20b · 95% interval 1280.9–1293.7 · 10,416 votes · 2026-10-28

Model performance

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

GPT OSS 20B

Tested configuration Radar task score

Unrated

1/4 core tests · not enough evidence

2 benchmark results for this configuration

Scientific reasoning

GPQA Diamond · 25%

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

GPT OSS 20B 60.8^*% Approx. 95% interval: 55.9–65.7% · 2026-08-27

Novel problem solving

Mystery Game Puzzles · 25%

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

GPT OSS 20B Not available

Advanced mathematics

FrontierMath Tiers 1-3 · 25%

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

GPT OSS 20B Not available

Factual answers

SimpleQA Verified · 25%

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

GPT OSS 20B Not available

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.

GPT OSS 20B 65.3^*% Approx. 95% interval: 53.7–76.9% · 2026-08-27

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.

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

GPT OSS 20B

1,314.72 rating

Confidence interval: 1,291.91 – 1,337.53

664 votes

Oct 28, 2026

Category: chinese
Evaluated configuration: gpt-oss-20b
Evaluation type: human preference

Arena text preference / codinghuman preference · 2026-10-28 · rating

GPT OSS 20B

1,306.41 rating

Confidence interval: 1,293.28 – 1,319.54

2112 votes

Oct 28, 2026

Category: coding
Evaluated configuration: gpt-oss-20b
Evaluation type: human preference

Arena text preference / creative_writinghuman preference · 2026-10-28 · rating

GPT OSS 20B

1,200.47 rating

Confidence interval: 1,182.69 – 1,218.24

1233 votes

Oct 28, 2026

Category: creative_writing
Evaluated configuration: gpt-oss-20b
Evaluation type: human preference

Arena text preference / englishhuman preference · 2026-10-28 · rating

GPT OSS 20B

1,302.82 rating

Confidence interval: 1,293.76 – 1,311.89

4513 votes

Oct 28, 2026

Category: english
Evaluated configuration: gpt-oss-20b
Evaluation type: human preference

Arena text preference / exclude_tieshuman preference · 2026-10-28 · rating

GPT OSS 20B

1,221.93 rating

Confidence interval: 1,213.03 – 1,230.84

7621 votes

Oct 28, 2026

Category: exclude_ties
Evaluated configuration: gpt-oss-20b
Evaluation type: human preference

Arena text preference / experthuman preference · 2026-10-28 · rating

GPT OSS 20B

1,257.79 rating

Confidence interval: 1,230.90 – 1,284.69

506 votes

Oct 28, 2026

Category: expert
Evaluated configuration: gpt-oss-20b
Evaluation type: human preference

Arena text preference / germanhuman preference · 2026-10-28 · rating

GPT OSS 20B

1,254.87 rating

Confidence interval: 1,214.50 – 1,295.24

217 votes

Oct 28, 2026

Category: german
Evaluated configuration: gpt-oss-20b
Evaluation type: human preference

Arena text preference / hard_promptshuman preference · 2026-10-28 · rating

GPT OSS 20B

1,273.65 rating

Confidence interval: 1,264.69 – 1,282.60

4718 votes

Oct 28, 2026

Category: hard_prompts
Evaluated configuration: gpt-oss-20b
Evaluation type: human preference

Arena text preference / hard_prompts_englishhuman preference · 2026-10-28 · rating

GPT OSS 20B

1,291.80 rating

Confidence interval: 1,278.48 – 1,305.11

2080 votes

Oct 28, 2026

Category: hard_prompts_english
Evaluated configuration: gpt-oss-20b
Evaluation type: human preference

Arena text preference / industry_business_and_management_and_financial_operationshuman preference · 2026-10-28 · rating

GPT OSS 20B

1,283.43 rating

Confidence interval: 1,269.40 – 1,297.46

1835 votes

Oct 28, 2026

Category: industry_business_and_management_and_financial_operations
Evaluated configuration: gpt-oss-20b
Evaluation type: human preference

Arena text preference / industry_entertainment_and_sports_and_mediahuman preference · 2026-10-28 · rating

GPT OSS 20B

1,215.94 rating

Confidence interval: 1,201.18 – 1,230.70

1728 votes

Oct 28, 2026

Category: industry_entertainment_and_sports_and_media
Evaluated configuration: gpt-oss-20b
Evaluation type: human preference

Arena text preference / industry_legal_and_governmenthuman preference · 2026-10-28 · rating

GPT OSS 20B

1,280.04 rating

Confidence interval: 1,256.66 – 1,303.42

678 votes

Oct 28, 2026

Category: industry_legal_and_government
Evaluated configuration: gpt-oss-20b
Evaluation type: human preference

Arena text preference / industry_life_and_physical_and_social_sciencehuman preference · 2026-10-28 · rating

GPT OSS 20B

1,296.04 rating

Confidence interval: 1,281.13 – 1,310.95

1669 votes

Oct 28, 2026

Category: industry_life_and_physical_and_social_science
Evaluated configuration: gpt-oss-20b
Evaluation type: human preference

Arena text preference / industry_mathematicalhuman preference · 2026-10-28 · rating

GPT OSS 20B

1,305.19 rating

Confidence interval: 1,281.03 – 1,329.36

585 votes

Oct 28, 2026

Category: industry_mathematical
Evaluated configuration: gpt-oss-20b
Evaluation type: human preference

Arena text preference / industry_medicine_and_healthcarehuman preference · 2026-10-28 · rating

GPT OSS 20B

1,309.73 rating

Confidence interval: 1,284.46 – 1,334.99

564 votes

Oct 28, 2026

Category: industry_medicine_and_healthcare
Evaluated configuration: gpt-oss-20b
Evaluation type: human preference

Arena text preference / industry_software_and_it_serviceshuman preference · 2026-10-28 · rating

GPT OSS 20B

1,311.01 rating

Confidence interval: 1,300.91 – 1,321.11

3591 votes

Oct 28, 2026

Category: industry_software_and_it_services
Evaluated configuration: gpt-oss-20b
Evaluation type: human preference

Arena text preference / industry_writing_and_literature_and_languagehuman preference · 2026-10-28 · rating

GPT OSS 20B

1,237.95 rating

Confidence interval: 1,224.94 – 1,250.95

2170 votes

Oct 28, 2026

Category: industry_writing_and_literature_and_language
Evaluated configuration: gpt-oss-20b
Evaluation type: human preference

Arena text preference / instruction_followinghuman preference · 2026-10-28 · rating

GPT OSS 20B

1,236.13 rating

Confidence interval: 1,223.91 – 1,248.35

2541 votes

Oct 28, 2026

Category: instruction_following
Evaluated configuration: gpt-oss-20b
Evaluation type: human preference

Arena text preference / japanesehuman preference · 2026-10-28 · rating

GPT OSS 20B

1,244.53 rating

Confidence interval: 1,205.99 – 1,283.08

246 votes

Oct 28, 2026

Category: japanese
Evaluated configuration: gpt-oss-20b
Evaluation type: human preference

Arena text preference / koreanhuman preference · 2026-10-28 · rating

GPT OSS 20B

1,233.88 rating

Confidence interval: 1,192.87 – 1,274.88

227 votes

Oct 28, 2026

Category: korean
Evaluated configuration: gpt-oss-20b
Evaluation type: human preference

Arena text preference / longer_queryhuman preference · 2026-10-28 · rating

GPT OSS 20B

1,250.04 rating

Confidence interval: 1,236.76 – 1,263.31

2103 votes

Oct 28, 2026

Category: longer_query
Evaluated configuration: gpt-oss-20b
Evaluation type: human preference

Arena text preference / mathhuman preference · 2026-10-28 · rating

GPT OSS 20B

1,316.89 rating

Confidence interval: 1,294.54 – 1,339.24

678 votes

Oct 28, 2026

Category: math
Evaluated configuration: gpt-oss-20b
Evaluation type: human preference

Arena text preference / multi_turnhuman preference · 2026-10-28 · rating

GPT OSS 20B

1,267.62 rating

Confidence interval: 1,253.13 – 1,282.12

1775 votes

Oct 28, 2026

Category: multi_turn
Evaluated configuration: gpt-oss-20b
Evaluation type: human preference

Arena text preference / non_englishhuman preference · 2026-10-28 · rating

GPT OSS 20B

1,268.39 rating

Confidence interval: 1,260.36 – 1,276.41

5892 votes

Oct 28, 2026

Category: non_english
Evaluated configuration: gpt-oss-20b
Evaluation type: human preference

Arena text preference / overallhuman preference · 2026-10-28 · rating

GPT OSS 20B

1,287.31 rating

Confidence interval: 1,280.92 – 1,293.70

10416 votes

Oct 28, 2026

Category: overall
Evaluated configuration: gpt-oss-20b
Evaluation type: human preference

Arena text preference / polishhuman preference · 2026-10-28 · rating

GPT OSS 20B

1,252.41 rating

Confidence interval: 1,230.43 – 1,274.39

809 votes

Oct 28, 2026

Category: polish
Evaluated configuration: gpt-oss-20b
Evaluation type: human preference

Arena text preference / russianhuman preference · 2026-10-28 · rating

GPT OSS 20B

1,277.96 rating

Confidence interval: 1,256.85 – 1,299.07

755 votes

Oct 28, 2026

Category: russian
Evaluated configuration: gpt-oss-20b
Evaluation type: human preference

Arena text preference / spanishhuman preference · 2026-10-28 · rating

GPT OSS 20B

1,267.47 rating

Confidence interval: 1,229.35 – 1,305.58

284 votes

Oct 28, 2026

Category: spanish
Evaluated configuration: gpt-oss-20b
Evaluation type: human preference

Run it locally

Inference · memory estimate

GPT OSS 20B

Not enough verified architecture information for a reliable VRAM estimate.

API cost

USD / month

Input tokens / monthOutput tokens / month

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

GPT OSS 20B

Provider offer

GPT OSS 20B / Pricing tiers 4

Provider / scope	Rate	Currency	Amount / unit
AWS Bedrockgpt-oss-20bStandard	Input	USD	0.07 / 1M tokens
AWS Bedrockgpt-oss-20bStandard Global cross-region	Input	USD	0.0721 / 1M tokens
AWS Bedrockgpt-oss-20bStandard Global cross-region	Output	USD	0.309 / 1M tokens
DeepInfraopenai/gpt-oss-20bStandard	Input	USD	0.03 / 1M tokens
DeepInfraopenai/gpt-oss-20bStandard	Output	USD	0.14 / 1M tokens
GroqGPT OSS 20BStandard	Input	USD	0.075 / 1M tokens
GroqGPT OSS 20BStandard	Output	USD	0.3 / 1M tokens

Capabilities

Context window: 131,072 tokens
Maximum output: 32,768 tokens
Reasoning: Yes
Tool calling: Yes
Open weights: Yes
Knowledge cutoff: 2024-06

Input modalities

text

Output modalities

text

Processing & platform

Architecture: gpt_oss

* Provisional data