Terminal-Bench leaderboard
- GPT-5.5 leads
- 84.7%
- models measured
- 40
- lead over #2
- +2.9 pts
- median
- 44.1%
Leader on Terminal-Bench · read October 22
# Terminal-Bench, read October 22
leaders = {
"GPT-5.5": 84.7%, "GPT-5.4": 81.8%, "Claude Opus 4.7": 80.2%,
}
On October 22, 2026, 40 of our catalog's models had a Terminal-Bench result, and OpenAI GPT-5.5 led with 84.7%, ahead of OpenAI GPT-5.4 at 81.8%.
Tasks an agent has to finish in a real terminal: compiling, configuring, debugging and data work. The number is the share of tasks solved, best agent per model, as Epoch AI collects it.
Role in the index
A coding board we show beside the Vibe Coding Index, not inside it: the index keeps the six boards it was defined on, so a score does not move every time a board is added. Read the methodology
| # | Model | Result | Evidence | Coding index | Output $/Mtok |
|---|---|---|---|---|---|
| 1 | GPT-5.5OpenAI | 84.7% | 71.3 | $30.00 | |
| 2 | GPT-5.4OpenAI | 81.8% | 64.4 | $15.00 | |
| 3 | Claude Opus 4.7Anthropic | 80.2% | 67.5 | $25.00 | |
| 4 | Gemini 3.1 Pro PreviewGoogle | 80.2% | 32.3 | $12.00 | |
| 5 | Claude Opus 4.6Anthropic | 79.8% | 57.2 | $25.00 | |
| 6 | GPT-5.3-CodexOpenAI | 78.4% | 60.0 | $14.00 | |
| 7 | GPT-5.2-CodexOpenAI | 66.5% | 45.4 | $14.00 | |
| 8 | GPT-5.2OpenAI | 64.9% | 55.9 | $14.00 | |
| 9 | Gemini 3 Flash PreviewGoogle | 64.3% | 43.7 | $3.00 | |
| 10 | Claude Opus 4.5Anthropic | 63.1% | 44.0 | $25.00 | |
| 11 | GPT-5.1-Codex-MiniOpenAI | 61.6% | Not benchmarked | $2.00 | |
| 12 | GPT-5.1-Codex-MaxOpenAI | 60.4% | Not benchmarked | $10.00 | |
| 13 | GPT-5.1-CodexOpenAI | 57.8% | 15.3 | $10.00 | |
| 14 | Grok 4.20xAI | 57.3% | 33.3 | $2.50 | |
| 15 | Claude Sonnet 4.6Anthropic | 53.4% | 57.3 | $15.00 | |
| 16 | GLM 5Z.ai | 52.4% | 41.7 | $3.20 | |
| 17 | GPT-5OpenAI | 49.6% | 38.8 | $10.00 | |
| 18 | GPT-5.1OpenAI | 47.6% | 40.4 | $10.00 | |
| 19 | Claude Sonnet 4.5Anthropic | 46.5% | 39.5 | $15.00 | |
| 20 | MiniMax M2.7MiniMax | 45.1% | 36.2 | $1.20 | |
| 21 | Kimi K2.5Moonshot AI | 43.2% | 38.6 | $2.85 | |
| 22 | MiniMax M2.5MiniMax | 42.7% | 37.2 | $1.20 | |
| 23 | DeepSeek V3.2DeepSeek | 39.6% | 30.0 | $0.88 | |
| 24 | Claude Opus 4.1Anthropic | 38.0% | Not benchmarked | $75.00 | |
| 25 | MiniMax M2.1MiniMax | 36.6% | Not benchmarked | $1.20 | |
| 26 | Kimi K2 ThinkingMoonshot AI | 35.7% | Not benchmarked | $2.50 | |
| 27 | Claude Haiku 4.5Anthropic | 35.5% | 35.6 | $5.00 | |
| 28 | GPT-5 MiniOpenAI | 34.8% | 34.2 | $2.00 | |
| 29 | GLM 4.7Z.ai | 33.4% | 55.1 | $2.20 | |
| 30 | Gemini 2.5 ProGoogle | 32.6% | 19.1 | $10.00 | |
| 31 | MiniMax M2MiniMax | 30.0% | Not benchmarked | $1.02 | |
| 32 | Kimi K2 0711Moonshot AI | 27.8% | Not benchmarked | $2.30 | |
| 33 | Qwen3 Coder 480B A35BQwen | 27.2% | Not benchmarked | $1.50 | |
| 34 | Qwen3.6 35B A3BQwen | 24.6% | Not benchmarked | $1.00 | |
| 35 | GLM 4.6Z.ai | 24.5% | 8.8 | $2.20 | |
| 36 | GPT-5 NanoOpenAI | 21.8% | Not benchmarked | $0.40 | |
| 37 | gpt-oss-120bOpenAI | 18.7% | Not benchmarked | $0.50 | |
| 38 | Gemini 2.5 FlashGoogle | 17.1% | Not benchmarked | $2.50 | |
| 39 | Qwen3.5-9BQwen | 9.2% | Not benchmarked | $0.15 | |
| 40 | gpt-oss-20bOpenAI | 3.4% | Not benchmarked | $0.15 |
Only models in our catalog are listed, and each appears once with its best published run. 215 rows were read from the source on October 22, 2026, and rows that match no catalog model are left out.