GPT-6 Sol
Category scores
Scores per capability area with the model's rank among ranked models in that category.
- 1Math97.3
- 2Multimodal86.6
- 3Knowledge79.9
- 4Coding#1277.5
- 5Agentic & tool use#3573.7
- 6Reasoning#1573.1
Benchmark results (24)
As reported in system cards, papers and public evaluations. Position compares this model with every other model reporting the same benchmark.
Agentic & tool use
| MazeBench | 7 | #10 of 42 |
| AutomationBench | 33.2 | #16 of 25 |
| OSWorld 2.0 | 60.5 | #7 of 24 |
| Agents' Last Exam | 56.4 | #2 of 17 |
| Hermes Index | 44.1 | #5 of 16 |
| Hermes Bench | 73.5 | #5 of 16 |
| ExploitGym | 22.1 | #5 of 14 |
| CWE-bench v1 | 52.0 | #10 of 11 |
Knowledge
| HealthBench Professional | 60.8 | #7 of 16 |
| HealthBench Hard | 30.1 | #9 of 12 |
| HealthBench (raw) | 47.1 | #7 of 7 |
| HealthBench (length-adjusted) | 53.2 | #7 of 7 |
| HealthBench Professional (raw) | 59.5 | #7 of 7 |
Reasoning
| ARC-AGI-2 | 89.6 | #8 of 41 |
| ARC-AGI-1 | 95.50 | #11 of 36 |
| ARC-AGI-3 | 4.6 | #7 of 22 |
| ARC-AGI-3 Provider Adapter | 23.03 | #5 of 8 |
Multimodal
| ParseBench | 69.69 | #15 of 37 |
| ExtractBench (LlamaIndex)(lower is better) | 94.57 | #23 of 24 |
Math
| FrontierMath v2 (Tiers 1-3) | 89.825 | #5 of 64 |
| FrontierMath v2 (Tier 4) | 90.000 | #5 of 51 |
Coding
| DeepSWE | 68.8 | #20 of 50 |
Instruction following
| Gray Swan IPI (15 attempts)(lower is better) | 10.1 | #7 of 12 |
Pricing by source
USD per 1M tokens.
| Source | Input | Output | Cached in |
|---|---|---|---|
| ✓ Official (OpenAI) checked Oct 17, 2026 · gpt-6-sol | $2.00 | $10.00 | $0.20 |
| BenchLM list price | $2.00 | $10.00 | $0.20 |
| OpenRouter openai/gpt-6-sol | $2.00 | $10.00 | $0.20 |
Standard tier, prompts under 272K tokens.
OpenAI's GPT-6 Sol model page lists $2.00 input / $0.20 cached input / $2.50 cache writes / $10.00 output per million tokens. Prompts above 272K input tokens are charged at 2x input and cache rates and 1.5x output for the full request; regional processing adds 10% where available; Batch and Flex are 50% of Standard and Fast mode is 2x.
Arena ratings
Human-preference ratings from LMArena. Best-rated variant shown.
| Text(max) | 1398 | #156 · 12.5K votes |
| WebDev (coding)(Max) | 1683 | #9 · 4.6K votes |
| Vision(Max) | 1240 | #74 · 3.2K votes |
| Search(max) | 1175 | #26 · 11.3K votes |
Speed
Measured by us through OpenRouter (median of trials).
Not measured yet.
Related models
Same family
- GPT-6 Astra84.0
- GPT-6 Luna66.6
Similar score
- Claude Opus 580.1 · $10.00
- Claude Fable 581.3 · $20.00
- GPT-5.6 Sol75.3 · $8.00
- Claude Sonnet 5.582.2 · $4.00
- Muse Wonder Full74.8 · —