OpenAI · Released Sep 4, 2026GPT-6 Astra
GPT-6 Astra is OpenAI's most capable model for complex reasoning, coding, computer use, research, and document creation.
- Capability
- 166.5ECI · #2 of 148
- Input
- $10.00per 1M tokens
- Output
- $50.00per 1M tokens
- Context
- 1.05M128K max output
What it costs
- Input
- $10.00per million tokens
- Output
- $50.00per million tokens
- Cached input
- $1.00per million tokens
- Cache write
- $12.50per million tokens
- Over 272K tokens
- $20.00 / $75.00input / output per million
- Blended (3:1)
- $20.00pricier than 96% of priced models
Typical monthly bills
| Monthly usage | Estimated cost |
|---|---|
| Side project2M input + 0.5M output tokens | $45.00 |
| Team assistant25M input + 5M output tokens | $500.00 |
| Production app250M input + 50M output tokens | $5,000 |
Official OpenAI API price, as listed on models.dev.
Independent benchmarks
#2 of 148 scored models
The shaded band is Epoch AI’s confidence range (163.1–171.1); the tick marks the median scored model.
Scores from Epoch AI, run independently of OpenAI.
The details
- Lab
- OpenAI
- Released
- Sep 4, 2026
- Knowledge cutoff
- Apr 30, 2026
- Context window
- 1,050,000 tokens
- Max output
- 128,000 tokens
- Inputs
- Text, Images, PDFs
- Output
- Text
- Reasoning
- Adjustable effort low · medium · high · xhigh · max
- Tool calling
- Yes
- Structured output
- Yes
- Weights
- Proprietary
- API model ID
gpt-6-astraon OpenAI- Availability
- 26 API providerslisted on models.dev
What OpenAI claims
Published by the lab at launch. Settings vary, so compare these only with care.
| Benchmark | Score | Setting | Source |
|---|---|---|---|
| Agents' Last Exam | 59.3 | — | Source |
| OSWorld v2.0 | 72.6 | — | Source |
| ScreenSpot-Pro | 92.7 | no tools | Source |
| AutomationBench | 41.4 success rate | — | Source |
| BenchCAD | 95.9 geometric overlap | with tools | Source |
| BrowseComp | 91.5 | — | Source |
| Terminal-Bench v4.0 | 57.9 success rate | — | Source |
| DeepSWE v1.1 | 74.1 resolve rate | — | Source |
| FrontierCode v1.1 | 64.5 | — | Source |
| Terminal-Bench Science v0.1 | 64.6 success rate | — | Source |
| FrontierMath vv2 | 97.6 | — | Source |
| GPQA Diamond | 96 | — | Source |
More from OpenAI
About GPT-6 Astra
How much does GPT-6 Astra cost?
GPT-6 Astra costs $10.00 input / $50.00 output per million tokens (official OpenAI API price). Cached input is $1.00 per million tokens, and writing to the cache costs $12.50. Requests over 272K tokens are billed at $20.00 input / $75.00 output. At a 3:1 input-to-output mix that is $20.00 per million tokens, more expensive than 96% of the 360 priced models we track.
What is the context window of GPT-6 Astra?
GPT-6 Astra accepts up to 1,050,000 tokens per request and can write up to 128,000 tokens in one response.
How good is GPT-6 Astra?
Epoch AI gives GPT-6 Astra a Capabilities Index score of 166.5 (likely range 163.1–171.1), ranking it #2 of 148 models Epoch has scored. Epoch AI benchmark results: GPQA Diamond 95.8%, FrontierMath Tiers 1–3 93.7%, OTIS Mock AIME 2024–2025 100%, SimpleQA Verified 75.6%.
Is GPT-6 Astra open source?
No. GPT-6 Astra is proprietary; you use it through OpenAI’s API or partner platforms.
What inputs does GPT-6 Astra support?
GPT-6 Astra accepts text, images and PDFs and replies in text. It is a reasoning model with low, medium, high, xhigh and max effort settings, supports tool calling and can return structured JSON output.
When was GPT-6 Astra released?
OpenAI released GPT-6 Astra on Sep 4, 2026. Its training data runs to Apr 30, 2026.
What are the best alternatives to GPT-6 Astra?
The closest current models from other labs on capability, price and release date are Claude Fable 5.1 (Anthropic, ECI 164.8, $10.00 / $50.00), Grok 4.7 (xAI, $2.00 / $6.00), Kimi K3 (Moonshot AI, ECI 157.6, $3.00 / $15.00) and Fugu Ultra (Sakana AI, $5.00 / $30.00).