Skip to content

Benchmark

Terminal-Bench 3.0

Agentic benchmarks measure how well models carry out tasks that take several steps. Scores are percentages. Higher is better.

Models measured
10
Scores recorded
21
Sources
2
IQ weight
No weight published

Score distribution

25.8%
Mean
28.3%
Median
9.5%
Std dev
14.9%
10th percentile
34.6%
90th percentile
7.6%
Min
7.6%43.3%
The band covers the middle 80% of scores. The line marks the median.

Leaderboard

#ModelScore
1Claude Opus 5DeepSeek Harness / max effort· Reported by DeepSeek· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.43.3%
2Claude Opus 5DeepSeek V4.1 launch chart · exact public value · setting not further specified· Reported by DeepSeek· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.43.3%
3GPT-5.6 Solsetting not named by the source· Reported by Z.ai· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.34.6%
4GPT-5.6 SolDeepSeek Harness / max effort· Reported by DeepSeek· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.34.4%
5GPT-5.6 SolDeepSeek V4.1 launch chart · exact public value · setting not further specified· Reported by DeepSeek· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.34.4%
6DeepSeek V4.1 Flash· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim30%
7DeepSeek V4.1 FlashDeepSeek Harness / max effort· Reported by DeepSeek· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.30%
8DeepSeek V4.1 FlashDeepSeek V4.1 launch chart · exact public value · setting not further specified· Reported by DeepSeek· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.30%
9GLM 5.3setting not named by the source· Reported by Z.ai· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.28.3%
10GLM 5.3· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim28.3%
11GLM 5.3DeepSeek Harness / max effort· Reported by DeepSeek· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.28.3%
12GLM 5.3DeepSeek V4.1 launch chart · exact public value · setting not further specified· Reported by DeepSeek· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.28.3%
13Grok 4.6· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim26%
14Claude Opus 4.8· Reported by Z.ai· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.21.1%
15Kimi K3DeepSeek Harness / max effort· Reported by DeepSeek· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.17.7%
16Kimi K3DeepSeek V4.1 launch chart · exact public value · setting not further specified· Reported by DeepSeek· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.17.7%
17Kimi K3setting not named by the source· Reported by Z.ai· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.17.4%
18Gemini 3.7 Flash· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim14.9%
19Gemini 3.7 Flashmini-swe-agent · highest thinking· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.14.9%
20DeepSeek V4 Pro 0813DeepSeek V4.1 launch chart · exact public value · setting not further specified· Reported by DeepSeek· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.11.8%
21DeepSeek V4 Flash 0731DeepSeek V4.1 launch chart · exact public value · setting not further specified· Reported by DeepSeek· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.7.6%

Data credits: BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links. · Data by LLM Stats