Chess with an engine
loads medium
Tech. quality98
UX92
Time10 min 32 s
Cost$2.26
Compare →
GPT-6 Astra
Loads without errors5/5#1 of 26
Tech. quality99/100#1 of 26
UX79/100#21 of 26
Total cost$7.65
Average time7 min 56 s#22 of 26
Total tokens1,191,837
API price: $10 / $50 per 1M input / output tokens · as of 15/10/2026 source ↗
What is GPT-6 Astra good for?
OpenAI's most capable model, the flagship for complex reasoning, coding and the most demanding work.
Description based on the maker · source ↗
- Maker
- OpenAI
- Released
- 03/09/2026
- Context window
- 1.05M tokens
- Model weights
- closed, the model is available through the API only
- Reasoning levels
- low medium high
How did GPT-6 Astra do in the AI Lab tests?
- 5 of 5 outputs load without errors.
- Its best technical result was on the “Solar System” prompt (100/100).
- It implemented the chess rules flawlessly – passing the objective perft test on all 6 positions.
- Average time per prompt 7 min 56 s#22 of 26
- Average cost per prompt $1.53#24 of 26
- Technical quality 99/100#1 of 26
- Automatic UX rating 79/100#21 of 26
Outputs
All outputs of GPT-6 Astra on the 5 prompts every other model got as well.
Solar System
loads medium
Tech. quality100
UX76
Time4 min 25 s
Cost$1.04
Compare →
GTA-style game
loads medium
Tech. quality100
UX77
Time11 min 09 s
Cost$2.06
Compare →
Monitor test
loads medium
Tech. quality100
UX89
Time4 min 18 s
Cost$0.78
Compare →
New York & Central Park map
loads medium
Tech. quality99
UX62
Time9 min 17 s
Cost$1.50
Compare →
Cost = tokens priced with the API price list valid on the date shown next to the cost (prices change). ≈ = estimate (only the total token count is known), no ≈ = exact from input / output tokens. “Loads” = the page opens without errors; technical quality also checks behaviour during use – judge the quality in the viewer.