0

GPT-6 Astra

Restricted

OpenAI's GPT-6 flagship, launched as its best computer-use model: it drives browsers, spreadsheets and desktop applications rather than answering questions about them. Built on OpenAI's largest training run to date and the first model it designates Critical for cybersecurity capability. Rolling out from Daybreak enterprise access to ChatGPT and the API; Pro, Business and Enterprise plans also get a GPT-6 Astra Pro tier.

Open
Publisher
OpenAI
Family
GPT
Context
1.1M
Input
Output
Providers
openai
Released
3 Sep 2026
Canonical
openai.com/index/gpt-6-astra

Cite

Notes

Only stored in your browser.

Attribution

Attribution policy →
Context
1.1M
Weights
Closed
API only
Input
$10
/1M tok
Output
$50
/1M tok
Blended
$20
#492/ 502/1M · 3:1
Speed
72
#81/ 501tok/s
Cache read
$1
0.10x input
Cache write
$12.50
1.25x input

Price & speed via Artificial Analysis · ranked across tracked models · caching via OpenRouter

Reasoning effortIndexSpeed
max52.772 t/s
xhigh52.465 t/s
high50.959 t/s
medium49.654 t/s
low45.856 t/s

Intelligence Index per reasoning-effort setting, via Artificial Analysis · pricing identical across tiers

Reported on 36 evals across 20 domains - best 100.0% on ExploitBench · #1 of 5

Reported eval scores

36
BrowseComp91.5
OSWorld72.6
Terminal-Bench57.9
ExploitBench100
GPQA Diamond96.1
SciCode56.5
EQ-Bench Creative Writing v32173.3
EQ-Bench Longform Creative Writing82.8
Creative Story-Writing Benchmark91
DeepSWE74.1
FrontierCode53.3
FrontierMath97.6
Agents' Last Exam59.3
Humanity's Last Exam (HLE)54.7
Vals Index63.1
ProofBench99
Vibe Code Bench v1.189.6
AutomationBench41.4
MedScribe87.9
MedCode48.5

Head-to-heads

29

benchmark comparisons against other models

vs GPT-5.519 shared evalsvs Claude Opus 4.819 shared evalsvs Claude Fable 518 shared evalsvs Gemini 3.1 Pro Preview16 shared evalsvs GLM 5.114 shared evalsvs Claude Opus 4.714 shared evalsvs GPT-5.213 shared evalsvs Claude Opus 4.613 shared evalsvs GPT-5.413 shared evalsvs Kimi K2.513 shared evalsvs Claude Sonnet 4.613 shared evalsvs GPT-5 Mini12 shared evalsvs GPT-512 shared evalsvs GPT-5.4 Mini12 shared evalsvs Claude Sonnet 4.511 shared evalsvs Claude Opus 4.511 shared evalsvs Qwen3.6 Plus11 shared evalsvs GPT-5 Nano10 shared evalsvs Gemini 2.5 Pro10 shared evalsvs Grok 4.1 Fast10 shared evalsvs DeepSeek V3.29 shared evalsvs GLM 4.79 shared evalsvs Claude 4 Sonnet8 shared evalsvs GPT-4.1 Mini6 shared evalsvs GPT-4.1 Nano6 shared evalsvs gpt-oss-120b6 shared evalsvs GPT-4.15 shared evalsvs GPT-4o-mini4 shared evalsvs GPT-4o4 shared evals