Benchmark
Terminal-Bench 2.1
Agentic benchmarks measure how well models carry out tasks that take several steps. Scores are percentages. Higher is better.
- Models and systems measured
- 64
- Scores recorded
- 159
- Sources
- 3
- IQ weight
- 2.76
Score distribution
- 75.3%
- Mean
- 82.7%
- Median
- 17.2%
- Std dev
- 56.4%
- 10th percentile
- 88.8%
- 90th percentile
- 2.9%
- Min
This benchmark has a weight of 2.76 for chat models, measured over 53 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.
Leaderboard
| # | Model or system | Score |
|---|---|---|
| 51 | Hy4 previewHF Eval Result· Reported by Tencent Hy· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 85.4% |
| 52 | GPT-5.6 SolAtria Dawn Preview official card / report Table 1 · published comparison run · model snapshots and reasoning tiers not specified unless stated · four-run mean · 12× timeout multiplier capped at 12 hours · 500 turns · Claude Code v2.1.224 except GPT models use Terminus 2· Reported by Shanghai AI Laboratory· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 85.1% |
| 53 | Claude Opus 4.8Z.ai复现· Reported by Z.ai· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 85% |
| 54 | DeepSeek V4.1 FlashOpenCode · max effort · temp 1.0 · top_p .95 · 1M · max_steps 500 · N=3 · no network· Reported by DeepSeek· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 85% |
| 55 | GPT-5.6 Luna· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 84.7% |
| 56 | GPT-5.6 LunaGemini 3.6 Flash Model Card · July 2026 · Terminus-2· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 84.7% |
| 57 | GPT-5.6 LunaOpenAI GPT-5.6 launch table· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 84.7% |
| 58 | Claude Fable 5Qwen表:Qwen 采用 Claude Code avg@10、5h、max_tokens 131072;友商取公开最高分· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 84.6% |
| 59 | Claude Opus 4.8Qwen表:Qwen 采用 Claude Code avg@10、5h、max_tokens 131072;友商取公开最高分· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 84.6% |
| 60 | Claude Fable 5· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 84.3% |
| 61 | Claude Fable 5Fable/Mythos 5 System Card capability summary · adaptive thinking · max effort · 5 trials unless noted· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 84.3% |
| 62 | Claude Fable 5max· Reported by SpaceXAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 84.3% |
| 63 | GLM 5.3 Flashsetting 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. | 84.3% |
| 64 | GLM 5.3 Flash· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 84.3% |
| 65 | DeepSeek V4.1 FlashCodex · max effort · temp 1.0 · top_p .95 · 1M · max_steps 500 · N=3 · no network· Reported by DeepSeek· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 84.1% |
| 66 | DeepSeek V4 Flash Vision Exp· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 83.9% |
| 67 | GPT 5.5Codex CLI· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 83.4% |
| 68 | GPT 5.5xhigh· Reported by SpaceXAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 83.4% |
| 69 | GPT 5.5xhigh · 89 tasks · bash-only · pass@1 · five-attempt mean· Reported by Meta, Meta Superintelligence Labs· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 83.4% |
| 70 | Grok 4.5· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 83.3% |
| 71 | Grok 4.5Gemini 3.6 Flash Model Card · July 2026 · Terminus-2· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 83.3% |
| 72 | Grok 4.5xAI launch comparison table· Reported by SpaceXAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 83.3% |
| 73 | Claude Fable 5OpenAI GPT-5.6 launch table· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 83.1% |
| 74 | Muse Spark 1.2· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 82.9% |
| 75 | Muse Spark 1.2Muse Code· Reported by Meta· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 82.9% |
| 76 | Muse Spark 1.2xhigh · 89 tasks · native coding harness · isolated cloud sandbox · pass@1· Reported by Meta· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 82.9% |
| 77 | Claude Opus 4.8Fable/Mythos 5 System Card capability summary · adaptive thinking · max effort · 5 trials unless noted· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 82.7% |
| 78 | Claude Opus 4.8max · 89 tasks · bash-only · pass@1 · five-attempt mean· Reported by Meta, Meta Superintelligence Labs· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 82.7% |
| 79 | DeepSeek V4 Flash 0731· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 82.7% |
| 80 | DeepSeek 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. | 82.7% |
| 81 | DeepSeek V4 Flash 0731official model-card setting· Reported by DeepSeek· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 82.7% |
| 82 | GLM 5.2· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 82.7% |
| 83 | GLM 5.2best reported harness · Claude Code 2.1.167 · 5-run mean · 128K max output· Reported by Z.ai· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 82.7% |
| 84 | Fugu UltraFugu official headline evaluation· Reported by Sakana AI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 82.1% |
| 85 | Ember-1· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 82% |
| 86 | GPT-5.6 TerraCodex · max· Reported by Meta· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 81.8% |
| 87 | Grok 4.5Grok Build· Reported by Meta· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 81.6% |
| 88 | Inkling 2 Small· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claim | 81.3% |
| 89 | Ling 3.1 Flash· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claim | 81.2% |
| 90 | GLM 5.2Terminus-2 · parser=json · 256K context · 4h timeout· Reported by Z.ai· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 81% |
| 91 | Claude Sonnet 5Gemini 3.6 Flash Model Card · July 2026 · Terminus-2· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 80.4% |
| 92 | Claude Sonnet 5mini-SWE-agent · xhigh · 89 tasks × 5 attempts· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 80.4% |
| 93 | Beam 1.0 501B A23B· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 80.1% |
| 94 | Muse Spark 1.1· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 80% |
| 95 | Muse Spark 1.1Meta Model API · xhigh · 89 tasks · bash-only · pass@1 · five-attempt mean· Reported by Meta, Meta Superintelligence Labs· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 80% |
| 96 | Claude Opus 4.7max· Reported by SpaceXAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 78.9% |
| 97 | Claude Opus 4.8max· Reported by SpaceXAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 78.9% |
| 98 | Claude Opus 4.8OpenAI GPT-5.6 launch table· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 78.9% |
| 99 | Gemini 3.6 FlashAntigravity CLI · high· Reported by Meta· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 78.9% |
| 100 | DeepSeek V4 Pro 0813Atria Dawn Preview official card / report Table 1 · published comparison run · model snapshots and reasoning tiers not specified unless stated · four-run mean · 12× timeout multiplier capped at 12 hours · 500 turns · Claude Code v2.1.224 except GPT models use Terminus 2· Reported by Shanghai AI Laboratory· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 78.7% |
Showing 51–100 of 158 results.
Not ranked
The source marks these results as descriptive, not capability scores, so we list them by name instead of ranking them.
| Model or system | Score |
|---|---|
| Muse Spark (version unspecified)Meta Model API · highest effort · version unspecified · 89 tasks · bash-only · pass@1 · five-attempt mean· Reported by Meta, Meta Superintelligence Labs· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 67.3% |
Data credits: Data by LLM Stats · Hugging Face official benchmark leaderboards; licence varies by benchmark · 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.