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 | |
|---|---|---|
| 201 | MiMo V2.6 Flashterminal-bench-4.0, mini-swe-agent· Terminal-Bench | 16.1% |
| 202 | MiMo V2.6 Flashterminal-bench-4.1, mini-swe-agent· Terminal-Bench | 16.1% |
| 203 | Grok 4.5terminal-bench-3.0, Cursor CLI· Terminal-Bench | 15.7% |
| 204 | Gemini 2.5 Flashterminal-bench-2.0, Gemini CLI· Terminal-Bench | 15.4% |
| 205 | GPT-6 Lunaterminal-bench-4.1, Codex· Terminal-Bench | 15.2% |
| 206 | Muse Spark 1.3terminal-bench-4.1, Muse Code· Terminal-Bench | 14.9% |
| 207 | Muse Spark 1.3terminal-bench-4.0, Muse Code· Terminal-Bench | 14.6% |
| 208 | GPT-5.6 Lunaterminal-bench-3.0, Codex· Terminal-Bench | 14.3% |
| 209 | GPT OSS 120Bterminal-bench-2.0, Mini-SWE-Agent· Terminal-Bench | 14.2% |
| 210 | Grok Code Fast 1terminal-bench-2.0, Terminus 2· Terminal-Bench | 14.2% |
| 211 | Claude Haiku 4.5terminal-bench-2.0, OpenHands· Terminal-Bench | 13.9% |
| 212 | Gemini 3.7 Flashterminal-bench-4.1, mini-SWE-agent· Terminal-Bench | 12.4% |
| 213 | Grok 4.5terminal-bench-4.0, Grok Build· Terminal-Bench | 12.4% |
| 214 | Grok 4.5terminal-bench-4.1, Grok Build· Terminal-Bench | 12.4% |
| 215 | Hy4 Flashterminal-bench-4.1, mini-swe-agent· Terminal-Bench | 11.8% |
| 216 | GPT-5 nanoterminal-bench-2.0, Codex CLI· Terminal-Bench | 11.5% |
| 217 | Gemini 3.7 Flashterminal-bench-4.0, mini-SWE-agent· Terminal-Bench | 11.2% |
| 218 | GPT-5 nanoterminal-bench-2.0, OpenHands· Terminal-Bench | 9.9% |
| 219 | Qwen3.5 9Bterminal-bench-2.0, little-coder· Terminal-Bench | 9.2% |
| 220 | GPT-5 nanoterminal-bench-2.0, Terminus 2· Terminal-Bench | 7.9% |
| 221 | MiniMax M3terminal-bench-4.1, mini-swe-agent· Terminal-Bench | 7.6% |
| 222 | MiniMax M3terminal-bench-4.0, mini-swe-agent· Terminal-Bench | 7.3% |
| 223 | GPT-5 nanoterminal-bench-2.0, Mini-SWE-Agent· Terminal-Bench | 7% |
| 224 | DeepSeek V4 Pro 0813terminal-bench-4.0, Claude Code· Terminal-Bench | 6.7% |
| 225 | DeepSeek V4 Pro 0813terminal-bench-4.1, Claude Code· Terminal-Bench | 6.7% |
| 226 | GLM 5.2terminal-bench-3.0, Claude Code· Terminal-Bench | 4.6% |
| 227 | GPT OSS 20bterminal-bench-2.0, Mini-SWE-Agent· Terminal-Bench | 3.4% |
| 228 | GPT OSS 20bterminal-bench-2.0, Terminus 2· Terminal-Bench | 3.1% |
| 229 | Gemini 3.1 Pro Previewterminal-bench-4.0, Antigravity CLI· Terminal-Bench | 2.7% |
| 230 | Gemini 3.1 Pro Previewterminal-bench-4.1, Antigravity CLI· Terminal-Bench | 2.7% |
| 231 | GPT-5.4 miniterminal-bench-4.0, Codex· Terminal-Bench | 1.8% |
| 232 | GPT-5.4 miniterminal-bench-4.1, Codex· Terminal-Bench | 1.8% |
| 233 | Inkling FP4terminal-bench-4.0, mini-swe-agent· Terminal-Bench | 0% |
| 234 | Inkling FP4terminal-bench-4.1, mini-swe-agent· Terminal-Bench | 0% |
Showing 201–234 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)