GPT-6 Astra
OpenAI · Released Sep 3, 2026
$10
Input price / MTok
#33 of 33$50
Output price / MTok
#33 of 331.1M
Context window
#4 of 3498.6
ARC-AGI-3 (abstract reasoning)
#1 of 197.6
FrontierMath Tier 4 v2
#1 of 1100
ExploitBench (cybersecurity)
#1 of 292.7
ScreenSpot-Pro (desktop automation)
#1 of 257.2
Humanity's Last Exam (with tools)
#1 of 161
Artificial Analysis Intelligence Index
#1 of 1Summary
GPT-6 Astra is a proprietary language model from OpenAI, released on Sep 3, 2026.
At $10 input and $50 output per million tokens, it ranks #33 of 33 for input price, placing it among the higher-priced models tracked here.
Its context window of 1.1M tokens ranks #4 of 34 for context depth in this dataset.
It targets agents, coding, research workloads.
Ships with web search, code interpreter, a hosted shell, apply-patch, computer use, and MCP support built in. Staged rollout starting with a limited partner preview Sep 3 2026, then ChatGPT Plus/Pro/Business/Enterprise plus the API and AWS over the following days. First OpenAI model to hit the 'Critical' cybersecurity threshold under the company's own Preparedness Framework; its most advanced cyber capabilities are gated behind the enterprise-only 'Daybreak' program rather than shipped broadly. Fortune reported OpenAI quietly revised some evaluation metrics upward after a delayed publication of the launch announcement, a credibility flag independent of the ARC-AGI-3 discrepancy above. Strongest, most independently-corroborated advantage across hands-on reviews is cost/token efficiency on agentic coding and robotics tasks rather than raw intelligence; the cross-vendor Artificial Analysis index and Humanity's Last Exam both still favor Claude Fable 5.1 and Opus 5. API identifier gpt-6-astra, knowledge cutoff Apr 30 2026.
Benchmarks
| Benchmark | Score | Max | % | Rank | Source |
|---|---|---|---|---|---|
| ARC-AGI-3 (abstract reasoning) | 98.6 | 100 | 98.6% | only entry | OpenAI (self-reported at launch, Sep 3 2... |
| FrontierMath Tier 4 v2 | 97.6 | 100 | 97.6% | only entry | OpenAI (self-reported at launch, Sep 3 2... |
| ExploitBench (cybersecurity) | 100 | 100 | 100.0% | #1/2 | OpenAI (self-reported at launch, Sep 3 2... |
| ScreenSpot-Pro (desktop automation) | 92.7 | 100 | 92.7% | #1/2 | OpenAI (self-reported at launch, Sep 3 2... |
| Humanity's Last Exam (with tools) | 57.2 | 100 | 57.2% | only entry | OpenAI (self-reported at launch, Sep 3 2... |
| Artificial Analysis Intelligence Index | 61 | 100 | 61.0% | only entry | Artificial Analysis (independent, cross-... |
Similar Models
Compare all 2 modelsModels with the most shared benchmark coverage and closest scores to GPT-6 Astra.
Compare GPT-6 Astra side by side
Pre-filled with 1 similar model