Open GPT reasoning model for self-hosted agents and controllable deployments
Specification
most-agreed values
Context
131K
Max output
33K
Released
2025-08-05
Knowledge cutoff
2024-06
Retires
—
Open weights
yes
Input
text
Output
text
Price
US dollars per million tokens · most-agreed
Input
free
Output
free
Cache read
$0.014
Quality
5 benchmarks · 3 sources
Composite
43rd
percentile of 344 scored models
Rank
197 / 344
±0.07 sd
Evidence
5 × 3
3 or more benchmarks, from 2 or more sources
Effort range
—
one configuration only
coding22nd2/3 bench
reasoning40th2/4 bench
math90th1/7 bench
Every score
one row per source and configuration — nothing averaged away
| Benchmark | Score | Configuration | Source | Run |
|---|---|---|---|---|
| OTIS Mock AIME 2024-2025 | 88.9 ±4.5 | high | Epoch AI | 2025-12-11 |
| GPQA diamond | 75.8 ±2.7 | high | Epoch AI | 2025-12-11 |
| Aider polyglot | 41.8 | high · aider diff | Aider | 2025-08-06 |
| Mystery Game Puzzles | 1.0 ±0.0 | default · 2 runs | Epoch AI | 2026-08-27 |
| SWE-Bench verified | 26.0 | default · mini-SWE-agent | SWE-bench | — |
The leading figure for a benchmark is the median across its configurations, so one heroic high-effort run cannot set the number.
Available from 5 hosts
| Host | In $/M | Out $/M | Cache rd | Cache wr | Context | Output | Retires |
|---|---|---|---|---|---|---|---|
| ollama · 120b-cloud | free | free | — | — | 131K | 131K | — |
| ollama · 20b-cloud | free | free | — | — | 131K | 131K | — |
| pendra · 120b | free | free | — | — | 131K | 33K | — |
| ollama-cloud · 20b | $0.07 | $0.30 | $0.035 | — | 131K | 33K | — |
| ollama-cloud · 120b | $0.15 | $0.60 | $0.014 | — | 131K | 33K | — |
Other listings of this model
same model, different key
More from OpenAI
most-hosted first