supermarket

Vision-language

GPT-6 Astra

By OpenAI, in the GPT-6 family.

Released
3 Sep 2026
Status
Current
Context
1.1M
Max output
128K
Input per 1M tokens
$10.00
Output per 1M tokens
$50.00
Open weights
No
License
proprietary
Parameters
—
Knowledge cutoff
30 Apr 2026
Reasoning
Yes
Modalities
text, image → text

Category scores

Out of 100, from independent and third-party results only, by method category-2026-09-29.1.

Category scores
CategoryScoreRankCoverage
Knowledge & reasoning74.223 of 4
Coding & software engineering58.023 of 5
Agentic tool use68.3Not ranked1 of 2
Vision & multimodal understanding71.4Not ranked1 of 3
Factuality & hallucination75.611 of 1
Human preference61.534 of 5

Benchmark results

Every result we hold for this model, with who ran it and where it was published.

Knowledge & reasoning

Knowledge & reasoning results
BenchmarkMetricValueProvenanceSettingEvaluatedSource
Chess PuzzlesPuzzles solved72.0%Independentmax30 Aug 2026Epoch AI
DTBench (Decision Theory Benchmark)Accuracy (capability questions)97.3%Third partyxhigh—https://conceptualreasoning.ai/dtbench via Epoch AI
DTBench (Decision Theory Benchmark)Accuracy (capability questions)97.1%Third partymedium—https://conceptualreasoning.ai/dtbench via Epoch AI
DTBench (Decision Theory Benchmark)Accuracy (capability questions)97.1%Third partyunknown—https://conceptualreasoning.ai/dtbench via Epoch AI
DTBench (Decision Theory Benchmark)Accuracy (capability questions)96.5%Third partyhigh—https://conceptualreasoning.ai/dtbench via Epoch AI
DTBench (Decision Theory Benchmark)Accuracy (capability questions)94.9%Third partylow—https://conceptualreasoning.ai/dtbench via Epoch AI
Earthborne Rangers (EBR-bench)Objectives reached76.2%Independentmax30 Aug 2026Epoch AI
LMCA (Language Model Conceptual Argumentation)Correlation with expert ratings (Pearson r)0.6437Third partyxhigh—https://conceptualreasoning.ai/lmca via Epoch AI
LMCA (Language Model Conceptual Argumentation)Correlation with expert ratings (Pearson r)0.6411Third partyunknown—https://conceptualreasoning.ai/lmca via Epoch AI
LMCA (Language Model Conceptual Argumentation)Correlation with expert ratings (Pearson r)0.6384Third partyhigh—https://conceptualreasoning.ai/lmca via Epoch AI
LMCA (Language Model Conceptual Argumentation)Correlation with expert ratings (Pearson r)0.6293Third partymedium—https://conceptualreasoning.ai/lmca via Epoch AI
LMCA (Language Model Conceptual Argumentation)Correlation with expert ratings (Pearson r)0.6219Third partylow—https://conceptualreasoning.ai/lmca via Epoch AI
Mystery Game PuzzlesPuzzles solved84.0%Independentmax30 Aug 2026Epoch AI

Scientific reasoning

Scientific reasoning results
BenchmarkMetricValueProvenanceSettingEvaluatedSource
GPQA DiamondAccuracy95.8%Independentmax30 Aug 2026Epoch AI

Mathematics

Mathematics results
BenchmarkMetricValueProvenanceSettingEvaluatedSource
FrontierMath Tier 4Accuracy97.6%Independenthigh30 Aug 2026Epoch AI
FrontierMath Tier 4Accuracy97.6%Independentxhigh30 Aug 2026Epoch AI
FrontierMath Tier 4Accuracy97.6%Independentmax30 Aug 2026Epoch AI
FrontierMath Tier 4Accuracy97.6%Independentmedium30 Aug 2026Epoch AI
FrontierMath Tier 4Accuracy87.8%Independentlow30 Aug 2026Epoch AI
FrontierMath Tier 4Accuracy82.9%Independentnone30 Aug 2026Epoch AI
FrontierMath Tiers 1-3Accuracy93.7%Independentmax30 Aug 2026Epoch AI
OTIS Mock AIME 2024-2025Accuracy100.0%Independentmax30 Aug 2026Epoch AI

Coding & software engineering

Coding & software engineering results
BenchmarkMetricValueProvenanceSettingEvaluatedSource
DeepSWEPass@174.1%Third partyxhigh—https://deepswe.datacurve.ai/ via Epoch AI
DeepSWEPass@173.2%Third partymax—https://deepswe.datacurve.ai/ via Epoch AI
DeepSWEPass@173.2%Third partyhigh—https://deepswe.datacurve.ai/ via Epoch AI
DeepSWEPass@172.8%Third partymedium—https://deepswe.datacurve.ai/ via Epoch AI
DeepSWEPass@167.0%Third partylow—https://deepswe.datacurve.ai/ via Epoch AI
FrontierCodeMain score (Mean@5)53.3%Third partymax—https://cognition.com/frontiercode via Epoch AI
MirrorCodeScore46.7%Independenthigh30 Aug 2026Epoch AI
WeirdML (v2)Average accuracy93.6%Third partypromax—https://htihle.github.io/weirdml.html via Epoch AI
WeirdML (v2)Average accuracy93.3%Third partymax—https://htihle.github.io/weirdml.html via Epoch AI
WeirdML (v2)Average accuracy92.9%Third partyhigh—https://htihle.github.io/weirdml.html via Epoch AI

Agentic tool use

Agentic tool use results
BenchmarkMetricValueProvenanceSettingEvaluatedSource
BALROGAverage progress68.3%Third partymax—https://balrogai.com/ via Epoch AI

Professional knowledge work

Professional knowledge work results
BenchmarkMetricValueProvenanceSettingEvaluatedSource
APEX-AgentsPass@1 (all rubric criteria met)64.7%Source not statedunknown—Epoch AI via Epoch AI

Vision & multimodal understanding

Vision & multimodal understanding results
BenchmarkMetricValueProvenanceSettingEvaluatedSource
Furniture AssemblyScore80.0%Independentmax10 Sep 2026Epoch AI

Factuality & hallucination

Factuality & hallucination results
BenchmarkMetricValueProvenanceSettingEvaluatedSource
SimpleQA VerifiedAccuracy (% correct)75.6%Independentmax30 Aug 2026Epoch AI

Human preference

Human preference results
BenchmarkMetricValueProvenanceSettingEvaluatedSource
Code Arena: WebDev (arena.ai)Arena score (Elo scale)1792Independentmax25 Sep 2026Arena
Code Arena: WebDev (arena.ai)Votes4908Independentmax25 Sep 2026Arena
Document Arena (arena.ai)Arena score (Elo scale)1473Independentmax13 Sep 2026Arena
Document Arena (arena.ai)Votes1621Independentmax13 Sep 2026Arena
Text Arena (arena.ai)Arena score (Elo scale)1478Independentmax25 Sep 2026Arena
Text Arena (arena.ai)Votes7185Independentmax25 Sep 2026Arena
Vision Arena (arena.ai)Arena score (Elo scale)1285Independentmax28 Sep 2026Arena
Vision Arena (arena.ai)Votes2869Independentmax28 Sep 2026Arena

Other GPT-6 versions

Sources