OpenAI · Closed weights · Released Sep 22, 2026 · 1 reasoning level
GPT-6 Sol
input $2.00/M · cached $0.20/M · cache write $2.50/M · output $10.00/M · list openai-gpt-6-sol-2026-09-26 (Sep 26, 2026) · cutoff declared Apr 20, 2026
Levels
Aggregate results at every reasoning level GPT-6 Sol was run under — score, behavior, tokens and cost per level.
| Level | Score | 95% int. | Rank | Acc. | Cov. | Halluc. | In | Cached | Reas. | Out | $/task | $/corr. | Run |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Medium | 50.3 | #2 of 4 | $0.0067 |
Token and cost columns are per-task averages, costs at list price; cached = cache read + write. Ranks are among all entries, not only this model. No latency: these runs used a discounted, slower service tier, so speed is not reported.
Fingerprint
Aggregate and axis scores against every measured entry. Ticks are the other entries; the colored marks are GPT-6 Sol, one per level.
Aggregate
Medium50.3 [48.3, 52.6] · #2 of 4
Knowledge
Medium19.8 [15.9, 24.3] · #2 of 4
Math
Medium80.8 ±1.5 · #2 of 4
Sub-families
Mediumreasoning level
| Sub-family | Score | 95% int. | Acc. | Cov. | Halluc. | $/task | Rank |
|---|---|---|---|---|---|---|---|
| Knowledge | |||||||
| Knowledge Depth | 19.8 | #2 | |||||
| Math | |||||||
| Ballpark | 80.8 | #2 | |||||