GPT-6 AstravsClaude Opus 5

SI Score, benchmarks, price and context compared, with every number sourced.

77.5
SI Score
#3 of 142 ranked models

OpenAI 100% confidence 100 percent, Full confidence

74.9
SI Score
#5 of 142 ranked models

Anthropic 100% confidence 100 percent, Full confidence

Pillar by pillar

GPT-6 AstraClaude Opus 5
87.1 Reasoning30% of score 84.3
92.9 Math15% of score 86.9
64.8 Coding40% of score 69.7
76.9 Preference15% of score 82.0

The basics

AttributeGPT-6 AstraClaude Opus 5
Input price, per 1M tokens $10.00OpenAI pricingOfficial Standard short-context rate; excludes Batch/Flex/cache discounts Retrieved Oct 9, 2026 · factual citation
Open source ↗ 		$5.00Anthropic API pricingOfficial Claude API base tokens; lowest short-context global Standard rate; cache, batch, fast and regional premiums excluded Retrieved Oct 9, 2026 · factual citation
Open source ↗
Output price, per 1M tokens $50.00OpenAI pricingOfficial Standard short-context rate; excludes Batch/Flex/cache discounts Retrieved Oct 9, 2026 · factual citation
Open source ↗ 		$25.00Anthropic API pricingOfficial Claude API base tokens; lowest short-context global Standard rate; cache, batch, fast and regional premiums excluded Retrieved Oct 9, 2026 · factual citation
Open source ↗
Context window 1.1Mmodels.devPublished source fact Retrieved Oct 9, 2026 · MIT
Open source ↗ 		1Mmodels.devPublished source fact Retrieved Oct 9, 2026 · MIT
Open source ↗
Released Sep 3, 2026OpenAI API changelogPublished source fact Retrieved Oct 9, 2026 · factual citation
Open source ↗ 		Jul 24, 2026models.devPublished source fact Retrieved Oct 9, 2026 · MIT
Open source ↗
Weights Closed Closed

Highlighted values are the lower price or the larger context window.

Shared benchmarks

29 in common. Best result ahead: GPT-6 Astra on 17, Claude Opus 5 on 8

Each side shows its best published result. Results can come from different settings or harnesses, so open a row to compare like with like before reading much into a small gap.

ARC-AGI-1 (public eval) 99% 99%

GPT-6 Astra

  • low effort98.3%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'low' reasoning effort. [variant] openai-gpt-6-astra-lowPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • medium effort98.8%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'medium' reasoning effort. [variant] openai-gpt-6-astra-mediumPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • high effort98.5%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'high' reasoning effort. [variant] openai-gpt-6-astra-highPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • xhigh effort99%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'xhigh' reasoning effort. [variant] openai-gpt-6-astra-xhighPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • max effort97.8%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'max' reasoning effort. [variant] openai-gpt-6-astra-maxPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗

Claude Opus 5

  • high effort99%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'high' output effort. [variant] anthropic-claude-opus-5-highPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • max effort99%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'max' output effort. [variant] anthropic-claude-opus-5-maxPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
About ARC-AGI-1 (public eval)
ARC-AGI-1 (semi-private) 98.5% 97.5%

GPT-6 Astra

  • low effort96.5%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'low' reasoning effort. [variant] openai-gpt-6-astra-lowPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • medium effort97.5%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'medium' reasoning effort. [variant] openai-gpt-6-astra-mediumPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • high effort98.5%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'high' reasoning effort. [variant] openai-gpt-6-astra-highPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • xhigh effort98.5%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'xhigh' reasoning effort. [variant] openai-gpt-6-astra-xhighPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • max effort97.5%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'max' reasoning effort. [variant] openai-gpt-6-astra-maxPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗

Claude Opus 5

  • high effort97.5%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'high' output effort. [variant] anthropic-claude-opus-5-highPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • max effort97.5%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'max' output effort. [variant] anthropic-claude-opus-5-maxPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
About ARC-AGI-1 (semi-private)
ARC-AGI-2 (public eval) 97.9% 97.1%

GPT-6 Astra

  • low effort94.2%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'low' reasoning effort. [variant] openai-gpt-6-astra-lowPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • medium effort96.7%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'medium' reasoning effort. [variant] openai-gpt-6-astra-mediumPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • high effort97.9%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'high' reasoning effort. [variant] openai-gpt-6-astra-highPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • xhigh effort97.9%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'xhigh' reasoning effort. [variant] openai-gpt-6-astra-xhighPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • max effort97.5%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'max' reasoning effort. [variant] openai-gpt-6-astra-maxPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗

Claude Opus 5

  • high effort93.2%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'high' output effort. [variant] anthropic-claude-opus-5-highPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • max effort97.1%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'max' output effort. [variant] anthropic-claude-opus-5-maxPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
About ARC-AGI-2 (public eval)
ARC-AGI-2 (semi-private) 95% 90.4%

GPT-6 Astra

  • low effort85.4%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'low' reasoning effort. [variant] openai-gpt-6-astra-lowPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • medium effort92.1%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'medium' reasoning effort. [variant] openai-gpt-6-astra-mediumPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • high effort92.1%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'high' reasoning effort. [variant] openai-gpt-6-astra-highPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • xhigh effort93.3%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'xhigh' reasoning effort. [variant] openai-gpt-6-astra-xhighPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • max effort95%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'max' reasoning effort. [variant] openai-gpt-6-astra-maxPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗

Claude Opus 5

  • high effort88.3%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'high' output effort. [variant] anthropic-claude-opus-5-highPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • max effort90.4%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'max' output effort. [variant] anthropic-claude-opus-5-maxPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
About ARC-AGI-2 (semi-private)
ARC-AGI-3 99.9% 30.2%

GPT-6 Astra

  • Published result99.9%Official model cards via models.devLab-reported; metric RHAE; transcribed by MIT models.dev catalog; not independently evaluated [variant] Responses APIPublished Sep 3, 2026 Retrieved Oct 9, 2026 · factual citation; MIT transcription
    Open source ↗

Claude Opus 5

  • high effort30.2%Official model cards via models.devLab-reported; metric RHAE; transcribed by MIT models.dev catalog; not independently evaluated [variant] high effortPublished Jul 24, 2026 Retrieved Oct 9, 2026 · factual citation; MIT transcription
    Open source ↗
About ARC-AGI-3
ARC-AGI-3 (semi-private) 62.7% 30.2%

GPT-6 Astra

  • low effort17.5%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'low' reasoning effort. [variant] openai-gpt-6-astra-lowPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • medium effort38.6%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'medium' reasoning effort. [variant] openai-gpt-6-astra-mediumPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • high effort54.8%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'high' reasoning effort. [variant] openai-gpt-6-astra-highPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • xhigh effort59.3%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'xhigh' reasoning effort. [variant] openai-gpt-6-astra-xhighPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
  • max effort62.7%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'max' reasoning effort. [variant] openai-gpt-6-astra-maxPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗

Claude Opus 5

  • high effort30.2%ARC PrizeARC Prize steward-published result; exact edition/subset; publication time is HTTP Last-Modified of the aggregate export, not the evaluation date. Evaluation conducted with 'high' output effort. [variant] anthropic-claude-opus-5-highPublished Oct 6, 2026 Retrieved Oct 9, 2026 · factual citation
    Open source ↗
About ARC-AGI-3 (semi-private)
FrontierMath Tier 4 (v2) 97.6% 73.2%

GPT-6 Astra

  • no reasoning82.9%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] nonePublished Aug 30, 2026 Retrieved Oct 9, 2026 · CC-BY
    Open source ↗
  • low effort87.8%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] lowPublished Aug 30, 2026 Retrieved Oct 9, 2026 · CC-BY
    Open source ↗
  • medium effort97.6%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] mediumPublished Aug 30, 2026 Retrieved Oct 9, 2026 · CC-BY
    Open source ↗
  • high effort97.6%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] highPublished Aug 30, 2026 Retrieved Oct 9, 2026 · CC-BY
    Open source ↗
  • xhigh effort97.6%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] xhighPublished Aug 30, 2026 Retrieved Oct 9, 2026 · CC-BY
    Open source ↗
  • max effort97.6%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] maxPublished Aug 30, 2026 Retrieved Oct 9, 2026 · CC-BY
    Open source ↗

Claude Opus 5

  • max effort73.2%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] maxPublished Jul 24, 2026 Retrieved Oct 9, 2026 · CC-BY
    Open source ↗
About FrontierMath Tier 4 (v2)
FrontierMath Tiers 1–3 (v2) 93.7% 85.6%

GPT-6 Astra

  • max effort93.7%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] maxPublished Aug 30, 2026 Retrieved Oct 9, 2026 · CC-BY
    Open source ↗

Claude Opus 5

  • max effort85.6%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] maxPublished Jul 24, 2026 Retrieved Oct 9, 2026 · CC-BY
    Open source ↗
About FrontierMath Tiers 1–3 (v2)
GPQA Diamond 96% 93.9%

GPT-6 Astra

  • max effort95.8%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] maxPublished Aug 30, 2026 Retrieved Oct 9, 2026 · CC-BY
    Open source ↗
  • Setting 296%Official model cards via models.devLab-reported; metric accuracy; transcribed by MIT models.dev catalog; not independently evaluated [variant] Published Sep 3, 2026 Retrieved Oct 9, 2026 · factual citation; MIT transcription
    Open source ↗

Claude Opus 5

  • low effort87.9%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] lowPublished Aug 6, 2026 Retrieved Oct 9, 2026 · CC-BY
    Open source ↗
  • max effort93.9%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] maxPublished Jul 24, 2026 Retrieved Oct 9, 2026 · CC-BY
    Open source ↗
  • Setting 392.9%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] Published Aug 6, 2026 Retrieved Oct 9, 2026 · CC-BY
    Open source ↗
About GPQA Diamond
Humanity's Last Exam (with tools) 57.2% 64.7%

GPT-6 Astra

  • Published result57.2%Official model cards via models.devLab-reported; metric accuracy; transcribed by MIT models.dev catalog; not independently evaluated [variant] with toolsPublished Sep 3, 2026 Retrieved Oct 9, 2026 · factual citation; MIT transcription
    Open source ↗

Claude Opus 5

  • Published result64.7%Official model cards via models.devLab-reported; metric accuracy; transcribed by MIT models.dev catalog; not independently evaluated [variant] with toolsPublished Jul 24, 2026 Retrieved Oct 9, 2026 · factual citation; MIT transcription
    Open source ↗
About Humanity's Last Exam (with tools)
LiveBench Coding: code completion 80.4% 82.6%

GPT-6 Astra

  • Published result80.4%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 9, 2026 · factual citation; Apache-2.0 code
    Open source ↗

Claude Opus 5

  • Published result82.6%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 9, 2026 · factual citation; Apache-2.0 code
    Open source ↗
About LiveBench Coding: code completion
LiveBench Coding: code generation 80.3% 80.3%

GPT-6 Astra

  • Published result80.3%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 9, 2026 · factual citation; Apache-2.0 code
    Open source ↗

Claude Opus 5

  • Published result80.3%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 9, 2026 · factual citation; Apache-2.0 code
    Open source ↗
About LiveBench Coding: code generation
LiveBench Coding: JavaScript 63.6% 77.3%

GPT-6 Astra

  • Published result63.6%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 9, 2026 · factual citation; Apache-2.0 code
    Open source ↗

Claude Opus 5

  • Published result77.3%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 9, 2026 · factual citation; Apache-2.0 code
    Open source ↗
About LiveBench Coding: JavaScript
LiveBench Coding: Python 65% 75%

GPT-6 Astra

  • Published result65%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 9, 2026 · factual citation; Apache-2.0 code
    Open source ↗

Claude Opus 5

  • Published result75%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 9, 2026 · factual citation; Apache-2.0 code
    Open source ↗
About LiveBench Coding: Python
LiveBench Coding: TypeScript 43.3% 43.3%

GPT-6 Astra

  • Published result43.3%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 9, 2026 · factual citation; Apache-2.0 code
    Open source ↗

Claude Opus 5

  • Published result43.3%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 9, 2026 · factual citation; Apache-2.0 code
    Open source ↗
About LiveBench Coding: TypeScript
LiveBench Math: AMPS Hard 98% 99.0%

GPT-6 Astra

  • Published result98%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 9, 2026 · factual citation; Apache-2.0 code
    Open source ↗

Claude Opus 5

  • Published result99.0%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 9, 2026 · factual citation; Apache-2.0 code
    Open source ↗
About LiveBench Math: AMPS Hard
LiveBench Math: competition math 97.1% 94.1%

GPT-6 Astra

  • Published result97.1%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 9, 2026 · factual citation; Apache-2.0 code
    Open source ↗

Claude Opus 5

  • Published result94.1%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 9, 2026 · factual citation; Apache-2.0 code
    Open source ↗
About LiveBench Math: competition math
LiveBench Math: integrals 100% 97%

GPT-6 Astra

  • Published result100%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 9, 2026 · factual citation; Apache-2.0 code
    Open source ↗

Claude Opus 5

  • Published result97%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 9, 2026 · factual citation; Apache-2.0 code
    Open source ↗
About LiveBench Math: integrals
LiveBench Math: olympiad 92.2% 92.8%

GPT-6 Astra

  • Published result92.2%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 9, 2026 · factual citation; Apache-2.0 code
    Open source ↗

Claude Opus 5

  • Published result92.8%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 9, 2026 · factual citation; Apache-2.0 code
    Open source ↗
About LiveBench Math: olympiad
LiveBench Math: simplify 70.5% 61.6%

GPT-6 Astra

  • Published result70.5%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 9, 2026 · factual citation; Apache-2.0 code
    Open source ↗

Claude Opus 5

  • Published result61.6%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 9, 2026 · factual citation; Apache-2.0 code
    Open source ↗
About LiveBench Math: simplify
LiveBench Reasoning: connections 100% 99.3%

GPT-6 Astra

  • Published result100%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 9, 2026 · factual citation; Apache-2.0 code
    Open source ↗

Claude Opus 5

  • Published result99.3%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 9, 2026 · factual citation; Apache-2.0 code
    Open source ↗
About LiveBench Reasoning: connections
LiveBench Reasoning: consecutive events 90.6% 77.6%

GPT-6 Astra

  • Published result90.6%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 9, 2026 · factual citation; Apache-2.0 code
    Open source ↗

Claude Opus 5

  • Published result77.6%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 9, 2026 · factual citation; Apache-2.0 code
    Open source ↗
About LiveBench Reasoning: consecutive events
LiveBench Reasoning: logic with navigation 88% 86%

GPT-6 Astra

  • Published result88%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 9, 2026 · factual citation; Apache-2.0 code
    Open source ↗

Claude Opus 5

  • Published result86%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 9, 2026 · factual citation; Apache-2.0 code
    Open source ↗
About LiveBench Reasoning: logic with navigation
LiveBench Reasoning: spatial 98% 100%

GPT-6 Astra

  • Published result98%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 9, 2026 · factual citation; Apache-2.0 code
    Open source ↗

Claude Opus 5

  • Published result100%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 9, 2026 · factual citation; Apache-2.0 code
    Open source ↗
About LiveBench Reasoning: spatial
LiveBench Reasoning: theory of mind 84.6% 78.8%

GPT-6 Astra

  • Published result84.6%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 9, 2026 · factual citation; Apache-2.0 code
    Open source ↗

Claude Opus 5

  • Published result78.8%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 9, 2026 · factual citation; Apache-2.0 code
    Open source ↗
About LiveBench Reasoning: theory of mind
LiveBench Reasoning: zebra puzzles 100% 100%

GPT-6 Astra

  • Published result100%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 9, 2026 · factual citation; Apache-2.0 code
    Open source ↗

Claude Opus 5

  • Published result100%LiveBenchLiveBench subtask result; dataset edition 2026_06_25; publication time is HTTP Last-Modified of the score CSV, not the dataset edition or individual evaluation date [variant] 2026_06_25Published Oct 7, 2026 Retrieved Oct 9, 2026 · factual citation; Apache-2.0 code
    Open source ↗
About LiveBench Reasoning: zebra puzzles
LMArena Text 1440.5 elo 1502.6 elo

GPT-6 Astra

  • Published result1440.5 eloLMArena / ArenaPublished source fact [variant] text / overallPublished Oct 8, 2026 Retrieved Oct 9, 2026 · CC-BY-4.0
    Open source ↗

Claude Opus 5

  • Published result1502.6 eloLMArena / ArenaPublished source fact [variant] text / overallPublished Oct 8, 2026 Retrieved Oct 9, 2026 · CC-BY-4.0
    Open source ↗
About LMArena Text
OTIS Mock AIME 2024–2025 100% 98.9%

GPT-6 Astra

  • max effort100%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] maxPublished Aug 30, 2026 Retrieved Oct 9, 2026 · CC-BY
    Open source ↗

Claude Opus 5

  • low effort93.3%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] lowPublished Aug 6, 2026 Retrieved Oct 9, 2026 · CC-BY
    Open source ↗
  • max effort98.9%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] maxPublished Jul 24, 2026 Retrieved Oct 9, 2026 · CC-BY
    Open source ↗
  • Setting 397.8%Epoch AI BenchmarkingEpoch-owned evaluation mean score; scores only, no benchmark questions [variant] Published Aug 6, 2026 Retrieved Oct 9, 2026 · CC-BY
    Open source ↗
About OTIS Mock AIME 2024–2025
Terminal-Bench 4.0 58.2% 53.9%

GPT-6 Astra

  • Setting 158.2%Terminal-BenchTerminal-Bench 4.0; published harness submission, 95% CI retained at source [variant] Codex; maxPublished Sep 10, 2026 Retrieved Oct 9, 2026 · Apache-2.0; factual citation
    Open source ↗
  • Setting 257.9%Official model cards via models.devLab-reported; metric success rate; transcribed by MIT models.dev catalog; not independently evaluated [variant] 4.0Published Sep 3, 2026 Retrieved Oct 9, 2026 · factual citation; MIT transcription
    Open source ↗
  • Setting 357.9%Terminal-BenchTerminal-Bench 4.0; published harness submission, 95% CI retained at source [variant] Codex; highPublished Sep 10, 2026 Retrieved Oct 9, 2026 · Apache-2.0; factual citation
    Open source ↗
  • Setting 457.9%Terminal-BenchTerminal-Bench 4.0; published harness submission, 95% CI retained at source [variant] Codex; xhighPublished Sep 10, 2026 Retrieved Oct 9, 2026 · Apache-2.0; factual citation
    Open source ↗
  • Setting 554.2%Terminal-BenchTerminal-Bench 4.0; published harness submission, 95% CI retained at source [variant] Codex; mediumPublished Sep 10, 2026 Retrieved Oct 9, 2026 · Apache-2.0; factual citation
    Open source ↗
  • Setting 650.6%Terminal-BenchTerminal-Bench 4.0; published harness submission, 95% CI retained at source [variant] Codex; lowPublished Sep 10, 2026 Retrieved Oct 9, 2026 · Apache-2.0; factual citation
    Open source ↗

Claude Opus 5

  • Setting 153.9%Terminal-BenchTerminal-Bench 4.0; published harness submission, 95% CI retained at source [variant] Claude Code; xhighPublished Sep 17, 2026 Retrieved Oct 9, 2026 · Apache-2.0; factual citation
    Open source ↗
  • Setting 251.8%Terminal-BenchTerminal-Bench 4.0; published harness submission, 95% CI retained at source [variant] Claude Code; maxPublished Sep 3, 2026 Retrieved Oct 9, 2026 · Apache-2.0; factual citation
    Open source ↗
  • Setting 350.3%Terminal-BenchTerminal-Bench 4.0; published harness submission, 95% CI retained at source [variant] Claude Code; highPublished Sep 17, 2026 Retrieved Oct 9, 2026 · Apache-2.0; factual citation
    Open source ↗
  • Setting 444.9%Terminal-BenchTerminal-Bench 4.0; published harness submission, 95% CI retained at source [variant] Claude Code; mediumPublished Sep 17, 2026 Retrieved Oct 9, 2026 · Apache-2.0; factual citation
    Open source ↗
  • Setting 534.9%Terminal-BenchTerminal-Bench 4.0; published harness submission, 95% CI retained at source [variant] Claude Code; lowPublished Sep 17, 2026 Retrieved Oct 9, 2026 · Apache-2.0; factual citation
    Open source ↗
About Terminal-Bench 4.0

Which should you choose?

  • For math, GPT-6 Astra leads by 6.0 points.
  • For coding, Claude Opus 5 leads by 4.9 points.
  • For preference, Claude Opus 5 leads by 5.1 points.
  • Claude Opus 5 costs less per input token ($5.00 vs $10.00 per 1M).
  • Confidence is 100% for GPT-6 Astra and 100% for Claude Opus 5; sources still to report can move either score.

These follow from the numbers above. They're not a verdict on your use case.

What changed

All releases · RSS feed

Alerts on this device

What to be alerted about
RSS feed