[whatmodel.sh](../../README.md) / GPT-6 Astra
# GPT-6 Astra
OpenAI · released 2026-09 · closed weights
**score 84.8** ±0.6 · #1 of 732 · from 41 benchmark results across 10 sources
Score within the [LMArena (text)](https://whatmodel.sh/?set=lmarena-text&score=set) comparison set only.
## What it's good at
● marks GPT-6 Astra among every model on that task or test, worst to best.
task vs the others rank counts tests
agent tasks ▁▁▃▆█▄▅▅▄▄▂● #6 of 313 24% 8
coding ▁▁▁▃▅▇█▇▅▄▃● #2 of 469 9.1% 3
chat & writing ▁▁▁▃▆▄▆█▆▇●▃ #29 of 402* 3.7% 1
math ▁▁▃▅█▆▄█▄▂▁● #1 of 431 26% 11
vision & documents ▁▁▂▂▁▃▄▂▆▇█● #12 of 153 5.5% 5
web search ▁▁▁▅▄█▃●·▁·▁ #10 of 84 5.2% 3
agent tasks · #6 of 313 · 24% of its score
Trails [Claude Sonnet 5.5](claude-sonnet-5-5.md), [Claude Fable 5.1](claude-fable-5-1.md), [GPT-6.1 Sol](gpt-6-1-sol.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
Arena Agent (October 2026 aggrega… 0.1296 IPS ▁▁▁▃▃▂█▁▂●▁▁ beats 55/59 avg -0.379 IPS 9.4% sharp
Terminal-Bench 4.0 (maker) 57.9% ▁·▁··▁▆▆▅█●▃ beats 29/33 very hard · avg 0.4% 5.2% sharp · at its level · small run
Humanity's Last Exam with tools (… 57.2% ▃·▃·▅····▅·● beats 7/8 very hard · avg 12% 2.2% at its level
BrowseComp (maker) 91.5% ▄··█▄··▄·▄█● beats 8/8 medium · avg 38% 1.7% typical
KlinikeBench strict pass@1 (333 i… 24.3% ▁▁·▁▁▁▁▅▃█▅● beats 25/27 very hard · avg 2.4% 1.6% at its level
DeepSWE v1.1 (resolved) (maker) 74.1% █··▃▃····▃·● beats 6/6 hard · avg 22% 1.5% typical
Thinkingbox-bench pass@1 (507 sta… 58.3% ▅··▃·▃·█▃▅●█ beats 11/15 very hard · avg 5.1% 1.3% sharp · at its level
Agents' Last Exam (metric/harness… 59.3% ▂·▂·█▄▂·▂▂▄● beats 14/14 very hard · avg 12% 1.3% at its level
coding · #2 of 469 · 9.1% of its score
Trails [Claude Opus 5.5](claude-opus-5-5.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
Arena WebDev (Code Arena) 1773.4 ▁▁▃▂▅█▃▆▅▅▁● beats 128/129 avg 981 6.5% sharp
DeepSWE v1.1 (resolved) (maker) 74.1% █··▃▃····▃·● beats 6/6 hard · avg 22% 1.5% typical
Luau Leetcode fully solved (Torpe… 87.0% ▁▁▁▁·▁▂▃▅▃█● beats 48/49 hard · avg 17% 1.2% sharp
chat & writing · #29 of 402 · 3.7% of its score
Trails [Gemini 3.6 Flash](gemini-3-6-flash.md), [GLM-5.3-Flash](glm-5-3-flash.md), [Gemini 3.5 Flash](gemini-3-5-flash.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
LMArena (text) 1475.6 ▁▁▁▂▃▃▃▆▆▇●▃ beats 354/385 avg 1260 3.7% blunt
math · #1 of 431 · 26% of its score
test result vs the others in its run difficulty counts why
FrontierMath tier 4 v2 97.6% ▁▂▁▂█▅▂▃▂▁●▁ beats 66/67 very hard · avg 0.1% 5.4% sharp
MathArena ArXivMath April 2026 (4… 90.8% ▁·····▅██▄·● beats 20/20 very hard · avg 7.6% 4.6% sharp
FrontierMath tiers 1–3 v2 93.7% ▂··▁▁▃▄▄█▆▄● beats 86/86 very hard · avg 1.2% 3.6% sharp · maxed out
MathArena ArXivMath June 2026 (48… 94.4% ▂···▄·██▆▆▄● beats 19/19 very hard · avg 5.1% 3.4% sharp
MathArena ArXivLean June 2026 (46… 65.2% ▄····▄▄███·● beats 9/9 very hard · avg 0.9% 3.1% sharp · at its level
MathArena BrokenArXiv April 2026 … 97.5% ▂▂▆█▃▃▂▂▂··● beats 18/18 very hard · avg 0.8% 2.3% sharp · noisy
MathArena ArXivMath May 2026 (40 … 95.0% ▁··▁▄█▄▁·▃·● beats 17/17 very hard · avg 11% 1.5% typical
MathArena ArXivMath August 2026, … 88.6% █▄·▄·······● beats 5/5 very hard · avg 5.9% 1.1% sharp
FrontierMath: Erdős problems 2.9% █··········● beats 6/6 very hard · avg 0.0% 0.7% sharp · too hard for it · noisy
MathArena BrokenArXiv August 2026… 81.9% ▄▄·····▄·▄·● beats 5/5 very hard · avg 6.9% 0.5% at its level · noisy
OTIS mock AIME 2024–2025 100.0% ▃▃▄▃▃▃▇▇█▃▂● beats 196/197 very hard · avg 7.2% 0.0% sharp · maxed out · noisy
vision & documents · #12 of 153 · 5.5% of its score
Trails [Muse Spark 1.2](muse-spark-1-2.md), [Claude Opus 4.7](claude-opus-4-7.md), [Gemini 3.7 Flash](gemini-3-7-flash.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
Arena vision (style controlled) 1284.8 ▁▁▂▂▂▂▃▄▅▅█● beats 136/152 avg 1061 2.4% blunt
Arena document (style controlled) 1472.6 ▁▁▃▃▄▃▇█●▇▄▅ beats 25/37 avg 1331 1.3% blunt
CharXiv (reasoning questions) 89.3% ▂▅▃▅▅▆▆▅▆█▆● beats 101/101 hard · avg 31% 1.0% small run
CharXiv descriptive questions 96.2% ▂▄▄▅▇▄██▇▂▁● beats 59/59 medium · avg 64% 0.5% maxed out · small run
ScreenSpot-Pro (maker) 92.7% ▂▃▅▆██▅▂···● beats 24/24 medium · avg 49% 0.3% blunt · small run
web search · #10 of 84 · 5.2% of its score
Trails [GPT-4o (2024-08-06)](gpt-4o-2024-08-06.md), [GPT-4.1 mini](gpt-4-1-mini.md), [GPT-4.1](gpt-4-1.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
Humanity's Last Exam with tools (… 57.2% ▃·▃·▅····▅·● beats 7/8 very hard · avg 12% 2.2% at its level
BrowseComp (maker) 91.5% ▄··█▄··▄·▄█● beats 8/8 medium · avg 38% 1.7% typical
Arena search systems (style contr… 1246.2 ▁·▁▁▂▂▂█▆▃▂● beats 34/36 avg 1069 1.3% blunt
other tests (13)
test result vs the others in its run difficulty counts why
EBR-bench (Epoch) 76.2% ▄▂·▆█▂█▆▄▆·● beats 24/25 very hard · avg 0.3% 8.7% sharp · at its level
Mystery Game Puzzles (Epoch) 84.0% █··▂▃▇▄█▂▄▂● beats 77/77 very hard · avg 9.4% 6.3% sharp · at its level · noisy
Furniture Assembly (Epoch) 80.0% █▆▆▆▆▂▆▂▂▂▂● beats 31/32 very hard · avg 1.2% 5.3% sharp · at its level
Chess Puzzles (Epoch) 76.0% █▁▁▁▁▂▃▃▃▂▁● beats 137/137 very hard · avg 5.8% 3.4% sharp · at its level · noisy
HealthBench Professional (maker) 64.7% ▄·▄··▄▄·▄█·● beats 8/8 hard · avg 20% 1.2% blunt · at its level
MirrorCode (Epoch) 60.0% ▄·█▄▄▄██·●▄█ beats 10/13 very hard · avg 1.6% 1.2% sharp · at its level · noisy
SimpleQA Verified 75.6% ▁▁▃▂▃▃▆▆█▄▂● beats 81/81 very hard · avg 8.0% 1.2% at its level
Icelandic LLM Leaderboard average… 70.3% ▁▁▃▂▂▄▁▃▄▄▄● beats 37/38 very hard · avg 14% 1.1% at its level
HealthBench Hard (maker) 34.2% ▂··▂····▂▄·● beats 7/8 very hard · avg 9.6% 0.9% blunt · at its level
HealthBench (maker) 56.9% ▃·▃·····█▃●▅ beats 6/8 hard · avg 24% 0.8% blunt · at its level
GPQA Diamond 95.8% ▁▁▁▁▂▄▆▄▅▇█● beats 211/211 hard · avg 39% 0.6% sharp · maxed out
RobloxQA v2.0 multiple choice (To… 98.5% ▁▁▃▁█▄▃▃▆▂▂● beats 47/49 easy · avg 77% 0.4% maxed out
Vectara hallucination leaderboard… 91.3% ▁▂▁▂▅█●▃▄▃▁▁ beats 58/98 very easy · avg 90% 0.0% blunt
source results share of the score
Epoch AI 11 36% ▇▇▇▇
LMArena leaderboard dataset 6 25% ▇▇▇
MathArena ArXivMath / BrokenArXiv / ArX… 7 17% ▇▇
OpenAI vendor-reported benchmark scores… 9 15% ▇▇
KlinikeBench: Evaluating Language Model… 1 2% ▇
Roblox LLM Leaderboard 2 2% ▇
Meridian Labs inspect_charxiv evaluatio… 2 2% ▇
One Success Isn't Reliability: Thinking… 1 1% ▇
Miðeind Icelandic LLM Leaderboard 1 1% ▇
Vectara hallucination leaderboard 1 0% ▇
prior: pull toward the middle 0% ▇
weighted average = 84.8
Licences: Epoch AI ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)) · LMArena leaderboard dataset ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)) · MathArena ArXivMath / BrokenArXiv / ArXivLean model-output datasets ([CC BY-SA 4.0](../../licenses/matharena-sharealike-NOTICE.txt)) · KlinikeBench: Evaluating Language Models Beyond Diagnostic Accuracy ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)) · Roblox LLM Leaderboard ([MIT](../../licenses/roblox-llm-leaderboard-MIT.txt)) · Meridian Labs inspect\_charxiv evaluation report ([MIT](https://github.com/meridianlabs-ai/inspect_charxiv/blob/317e4ff4335d7d4242d8e3c803898b7345973bd2/LICENSE)) · One Success Isn't Reliability: Thinkingbox ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)) · Miðeind Icelandic LLM Leaderboard ([CC BY 4.0](../../licenses/mideind-icelandic-llm-leaderboard-CC-BY-4.0.txt)) · Vectara hallucination leaderboard ([Apache-2.0](https://www.apache.org/licenses/LICENSE-2.0))
## How to run it
### Through an API
api (OpenAI) · $20.00/M blended · $10 in / $50 out per 1M tokens · [platform.openai.com](https://platform.openai.com/docs/models)
[price source](https://github.com/anomalyco/models.dev/blob/a668908cfe7073faa0363699486fc41a3ff52412/providers/openai/models/gpt-6-astra.toml): Standard base token rate from models.dev. Provider-specific tiers, long-context surcharges, caching and promotions may differ. · [context source](https://github.com/anomalyco/models.dev/blob/a668908cfe7073faa0363699486fc41a3ff52412/models/openai/gpt-6-astra.toml)
[interactive version](https://whatmodel.sh/model/gpt-6-astra) · [JSON](https://whatmodel.sh/api?format=json&all=1)