Benchmark
TBench
Agentic benchmarks measure how well models carry out tasks that take several steps. Scores are percentages.
- Models measured
- 89
- Scores recorded
- 234
- Sources
- 1
- IQ weight
- 0.93
Score distribution
- 42.9%
- Mean
- 42.2%
- Median
- 22.9%
- Std dev
- 14.2%
- 10th percentile
- 75.1%
- 90th percentile
- 0%
- Min
This benchmark has a weight of 0.93 for chat models, measured over 89 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.
Leaderboard
| # | Model | |
|---|---|---|
| 101 | GPT 5terminal-bench-2.0, OpenHands· Terminal-Bench | 43.8% |
| 102 | GPT-5 Codexterminal-bench-2.0, Terminus 2· Terminal-Bench | 43.4% |
| 103 | Kimi K2.5terminal-bench-2.0, Terminus 2· Terminal-Bench | 43.2% |
| 104 | Claude Sonnet 4.5terminal-bench-2.0, Goose· Terminal-Bench | 43.1% |
| 105 | GPT-5.1 Codex Miniterminal-bench-2.0, Crux· Terminal-Bench | 43.1% |
| 106 | MiniMax M2.7terminal-bench-2.0, Harness Agent· Terminal-Bench | 42.9% |
| 107 | Claude Sonnet 4.5terminal-bench-2.0, Terminus 2· Terminal-Bench | 42.8% |
| 108 | Claude Sonnet 4.5terminal-bench-2.0, MAYA-V2· Terminal-Bench | 42.7% |
| 109 | MiniMax M2.5terminal-bench-2.0, cchuter· Terminal-Bench | 42.7% |
| 110 | Claude Sonnet 4.5terminal-bench-2.0, OpenHands· Terminal-Bench | 42.6% |
| 111 | Claude Sonnet 4.5terminal-bench-2.0, Mini-SWE-Agent· Terminal-Bench | 42.5% |
| 112 | MiniMax M2.5terminal-bench-2.0, Terminus 2· Terminal-Bench | 42.2% |
| 113 | GLM 5.3terminal-bench-4.0, Claude Code· Terminal-Bench | 41.8% |
| 114 | GPT-5 Codexterminal-bench-2.0, Mini-SWE-Agent· Terminal-Bench | 41.3% |
| 115 | GLM 5.3terminal-bench-4.1, Claude Code· Terminal-Bench | 41.2% |
| 116 | Claude Sonnet 4.5terminal-bench-2.0, Claude Code· Terminal-Bench | 40.1% |
| 117 | DeepSeek V4.1 Flash 1017terminal-bench-4.0, Claude Code· Terminal-Bench | 40% |
| 118 | DeepSeek V3.2terminal-bench-2.0, Terminus 2· Terminal-Bench | 39.6% |
| 119 | Hy 4terminal-bench-4.0, Claude Code· Terminal-Bench | 39.1% |
| 120 | Grok 4.7terminal-bench-4.1, Grok Build· Terminal-Bench | 38.8% |
| 121 | DeepSeek V4.1 Flash 1017terminal-bench-4.1, Claude Code· Terminal-Bench | 38.5% |
| 122 | GPT-5.6 Solterminal-bench-4.1, Codex· Terminal-Bench | 38.5% |
| 123 | Claude Opus 4.1terminal-bench-2.0, Terminus 2· Terminal-Bench | 38% |
| 124 | Hy 4terminal-bench-4.1, Claude Code· Terminal-Bench | 37.9% |
| 125 | Grok 4.7terminal-bench-4.0, Grok Build· Terminal-Bench | 37.6% |
| 126 | GPT-5.6 Solterminal-bench-4.0, Codex· Terminal-Bench | 37.3% |
| 127 | GLM 5.3 Flashterminal-bench-4.1, Claude Code· Terminal-Bench | 37% |
| 128 | Claude Opus 4.1terminal-bench-2.0, OpenHands· Terminal-Bench | 36.9% |
| 129 | GPT-5.1 Codexterminal-bench-2.0, Terminus 2· Terminal-Bench | 36.9% |
| 130 | GLM 5.3 Flashterminal-bench-4.0, Claude Code· Terminal-Bench | 36.7% |
| 131 | MiniMax M2.1terminal-bench-2.0, Crux· Terminal-Bench | 36.6% |
| 132 | Kimi K2 Thinkingterminal-bench-2.0, Terminus 2· Terminal-Bench | 35.7% |
| 133 | Claude Haiku 4.5terminal-bench-2.0, Goose· Terminal-Bench | 35.5% |
| 134 | GPT 5terminal-bench-2.0, Terminus 2· Terminal-Bench | 35.2% |
| 135 | Claude Opus 4.1terminal-bench-2.0, Mini-SWE-Agent· Terminal-Bench | 35.1% |
| 136 | Claude Opus 4.1terminal-bench-2.0, Claude Code· Terminal-Bench | 34.8% |
| 137 | GPT-5 miniterminal-bench-2.0, spoox-o-m· Terminal-Bench | 34.8% |
| 138 | GPT-5.6 Solterminal-bench-3.0, Codex· Terminal-Bench | 34.6% |
| 139 | GPT 5terminal-bench-2.0, Mini-SWE-Agent· Terminal-Bench | 33.9% |
| 140 | GLM 4.7terminal-bench-2.0, Terminus 2· Terminal-Bench | 33.4% |
| 141 | GLM 4.7terminal-bench-2.0, Crux· Terminal-Bench | 33.3% |
| 142 | Gemini 2.5 Proterminal-bench-2.0, Terminus 2· Terminal-Bench | 32.6% |
| 143 | GLM 5.3terminal-bench-3.0, Claude Code· Terminal-Bench | 32.4% |
| 144 | GPT-5 miniterminal-bench-2.0, Codex CLI· Terminal-Bench | 31.9% |
| 145 | Muse Emberterminal-bench-4.1, Muse Code· Terminal-Bench | 30.6% |
| 146 | MiniMax M2terminal-bench-2.0, Terminus 2· Terminal-Bench | 30% |
| 147 | Nemotron 3.5 Super VLterminal-bench-4.0, mini-swe-agent· Terminal-Bench | 30% |
| 148 | Nemotron 3.5 Super VLterminal-bench-4.1, mini-swe-agent· Terminal-Bench | 30% |
| 149 | Claude Haiku 4.5terminal-bench-2.0, Mini-SWE-Agent· Terminal-Bench | 29.8% |
| 150 | GPT-5 miniterminal-bench-2.0, OpenHands· Terminal-Bench | 29.2% |
Showing 101–150 of 234 results.
Tokens the model generated per task, as measured by the source. Comparable within this benchmark only.
Data credits: Terminal-Bench (Laude Institute), Apache-2.0 (licence)