GPT-6.1 Sol
Near-Astra intelligence at less than a quarter of Astra's measured task cost. This is the closed model to try first.
Every published effort
| Setting | Index | Per task | Tokens / sec |
|---|---|---|---|
| Max | 52 | $0.72 | 65 |
| Extra high | 51 | $0.39 | 63 |
| High | 50 | $0.32 | 63 |
| Medium | 48 | $0.21 | 59 |
| Low | 42 | $0.13 | 60 |
What this workload costs
List price for one call. Reasoning tokens count as output. A cache hit uses the lab’s cache rate when one is published.
Specs
- API id
- gpt-6.1-sol
- Released
- 29 Sep 2026
- Context
- 1M
- Max output
- —
- Input / 1M
- $2.00
- Output / 1M
- $10.00
- Cache read / 1M
- $0.10
- Knowledge
- —
- Inputs
- Text
- Weights
- Closed
Replaces GPT-6 Sol after seven days, at the same $2 / $10 sticker. Cache reads are 95% off, so $0.10. The same 272k whole-request reprice as Astra applies. Context figure is the 1M listed by Artificial Analysis.
Image input is confirmed on the Astra card. It is not repeated on the pricing row for Sol, so this card does not assume vision.
Use it when
The default OpenAI model. Coding and agent work at a price you can leave on. High effort is 50 on the index at $0.32 a measured task. Max is 52 at $0.72.
Leave it when
When you have measured that Astra's extra tools or its last point actually change the outcome. Otherwise Sol is the better bill.