Leaderboard
Terminal-Bench Hard
22 models scored · metric: success rate. Prices are the cheapest listed across all serving providers. Pts/$ = score ÷ blended price (3×input + output) / 4 — how much score each dollar buys.
Best value · top 5 by pts per dollar
- 1.Step 3.5 Flash 2603217
- 2.Step 3.5 Flash192
- 3.Step 3.7 Flash85.5
- 4.Mistral Small 466.3
- 5.GLM-4.5-Air48.2
| # | Model | Score | Best in /M | Best out /M | Pts / $ | Context |
|---|---|---|---|---|---|---|
| 🥇 | GPT-5-Codexopenai | 37.9 | $1.10 | $9 | 12.3 | 400K |
| 🥈 | Step 3.7 Flashstepfun | 35.6 | $0.185 | $1.11 | 85.5 | 256K |
| 🥉 | Step 3.5 Flash 2603stepfun | 32.6 | $0.1 | $0.3 | 217 | 256K |
| 4 | Step 3.5 Flashstepfun | 27.3 | $0.09 | $0.3 | 192 | 256K |
| 5 | Gemini 2.5 Progoogle | 26.5 | $0.87 | $7 | 11.0 | 1.05M |
| 6 | GLM-4.6zhipuai | 25 | $0.6 | $2.20 | 25.0 | 205K |
| 7 | GLM-4.5zhipuai | 22 | $0.6 | $2.20 | 22.0 | 131K |
| 8 | Qwen3 Maxalibaba | 20.5 | $1.20 | $6 | 8.5 | 262K |
| 9 | GLM-4.5-Airzhipuai | 20.5 | $0.2 | $1.10 | 48.2 | 131K |
| 10 | Devstral 2mistral | 18.9 | $0.4 | $2 | 23.6 | 262K |
| 11 | Mistral Small 4mistral | 17.4 | $0.15 | $0.6 | 66.3 | 256K |
| 12 | Mistral Large 3mistral | 15.9 | $0.5 | $1.50 | 21.2 | 262K |
| 13 | Qwen3-Coder 30B-A3B Instructalibaba | 15.2 | $0.45 | $2.25 | 16.9 | 262K |
| 14 | Gemini 2.5 Flashgoogle | 13.6 | $0.21 | $1.80 | 22.4 | 1.05M |
| 15 | GPT-4o (2024-08-06)openai | 8.3 | $2.50 | $10 | 1.9 | 128K |
| 16 | GPT-4o (2024-11-20)openai | 8.3 | $2.50 | $10 | 1.9 | 128K |
| 17 | DeepSeek-R1deepseek | 6.1 | $0.7 | $2.50 | 5.3 | 128K |
| 18 | Mistral Large 2.1mistral | 6.1 | $2 | $6 | 2.0 | 131K |
| 19 | GLM-4.5Vzhipuai | 5.3 | $0.6 | $1.80 | 5.9 | 64K |
| 20 | Gemini 2.5 Flash-Litegoogle | 4.5 | $0.1 | $0.1 | 45.0 | 1.05M |
| 21 | Mistral Medium 3mistral | 3.8 | $0.4 | $2 | 4.7 | 131K |
| 22 | Llama-3.3-70B-Instructmeta | 3 | $0.22 | $0.5 | 10.3 | 128K |
Cheapest scorer: Step 3.5 Flash at $0.09 input /M.