Models › Text › GPT-6 Astra
OpenAI · Model details, evidence, and API facts
GPT-6 Astra is OpenAI’s flagship model for end-to-end professional work
Starting provider API rate; context tiers apply
See what GPT-6 Astra does best, then review its limits, benchmarks, and pricing below
GPT-6 Astra handles computer use, coding, research, and document creation
Every GPT-6 Astra result below comes from a named source and is labeled by protocol, so each score reflects how it was measured
| Benchmark | Score | Evaluation details | Source |
|---|---|---|---|
AutomationBench-AA Tests whether GPT-6 Astra completes multi-step workflows across simulated SaaS apps in six domains. |
69% Guardrail-aware score |
Version: AutomationBench-AA 2026 • Split: 657 tasks across Finance, HR, Marketing, Operations, Sales, and Support • Harness: Artificial Analysis independent implementation • Attempts: One evaluated run per task • Evaluator: Programmatic end-state grader with guardrail checks • Reasoning effort: max | Methodology Artificial Analysis Independently reproduced |
DeepSWE 1.1 Measures GPT-6 Astra on 113 long-horizon software-engineering tasks graded by hand-written tests. |
74%±3 Pass@1 |
Version: DeepSWE 1.1 • Split: 113 tasks across 91 repositories • Harness: mini-swe-agent • Coverage: five languages • Confidence interval: ±3 | Methodology DataCurve Benchmark owner |
GPQA Diamond Tests GPT-6 Astra on 198 Google-proof graduate-level questions in biology, chemistry, and physics to probe expert-level scientific reasoning. |
96.0% Accuracy |
Version: GPQA Diamond 198 • Split: 198 Diamond questions • Harness: OpenAI launch evaluation table • Reasoning effort: max | Methodology OpenAI Provider-reported |
Humanity's Last Exam Tests GPT-6 Astra on 2,500 expert-level questions across dozens of academic fields, with tool use enabled. |
57.2% Accuracy |
Version: Humanity's Last Exam • Protocol: Tool-assisted, reported separately from no-tools results • Harness: OpenAI launch evaluation table • Score basis: Maximum result at any reasoning effort | Methodology OpenAI Provider-reported |
Terminal-Bench 4.0 Tests GPT-6 Astra on complex terminal-based tasks spanning software engineering, system configuration, and data analysis, rewarding long-horizon tool use and error recovery. |
59% Independent task success rate |
Version: Terminal-Bench 4.0 • Harness: Artificial Analysis evaluation • Index: Intelligence Index component • Reasoning effort: max | Methodology Artificial Analysis Independently reproduced |
Arena rankings capture how people vote between anonymous responses. GPT-6 Astra’s human preference standing is reported per leaderboard, since results can vary widely by task
GPT-6 Astra ranked first on Arena’s WebDev leaderboard for web apps in September, scoring 1800 from 2,281 votes
GPT-6 Astra offers a 1,050,000-token context window, 128,000 output tokens, image input, and adjustable reasoning effort, with full API pricing and tiers shown below
| Model | |
| Provider | OpenAI |
| Release date | Sep 3, 2026 |
| Catalog added | Sep 22, 2026 |
| Context window | 1,050,000 tokens |
| Maximum output | 128,000 tokens |
| Knowledge cutoff | 2026-04-30 |
| License | Proprietary |
| Capabilities and API | |
| Input types | Image, Text |
| Output types | Text |
| Reasoning | Supported (low to max effort) |
| Provider endpoint tool parameters | Accepts tool parameters |
| Base provider API input | $10.00 / 1M tokens |
| Base provider cached input | $1.00 / 1M tokens |
| Base provider API output | $50 / 1M tokens |
| Provider tier above 272,000 prompt tokens | Input $20.00 / 1M • Output $75 / 1M • Cached input $2.00 / 1M |
Provider token rates are context-dependent. The base rates apply below the listed prompt thresholds; the matching tier applies at or above each threshold. These are reference provider API rates, not Pyxa AI checkout charges — access in Pyxa AI follows the allowances of your Pyxa AI plan.
Each GPT-6 Astra benchmark and specification links to its original source, so you can check where every figure comes from
Discover where GPT-6 Astra performs best, from agentic coding and research to computer use and professional document creation
Put GPT-6 Astra to work on your coding, research, and document tasks