Benchmark
Terminal-Bench Science 0.1
Agentic benchmarks measure how well models carry out tasks that take several steps. Scores are percentages. Higher is better.
- Models and systems measured
- 19
- Scores recorded
- 53
- Sources
- 2
- IQ weight
- 2.76
Score distribution
- 45.1%
- Mean
- 52.6%
- Median
- 18.5%
- Std dev
- 21%
- 10th percentile
- 64.6%
- 90th percentile
- 9.2%
- Min
This benchmark has a weight of 2.76 for chat models, measured over 15 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 |
|---|---|---|
| 1 | GPT-6.2 Astra· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 77.1% |
| 2 | Claude Fable 5.2· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 74.6% |
| 3 | GPT-6 AstraTerminal-Bench Science 0.1 · OpenAI release chart · max reasoning · displayed cost $23.80/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 68.1% |
| 4 | GPT-6.2 Sol· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 68% |
| 5 | Gemini 4 Carbon· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 64.8% |
| 6 | GPT-6 Astra· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 64.6% |
| 7 | GPT-6 Astrahighest effort· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 64.6% |
| 8 | GPT-6 AstraOpus 5.5 System Card capability section · OpenAI-reported · max effort· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 64.6% |
| 9 | GPT-6 AstraTerminal-Bench Science 0.1 · OpenAI release chart · high reasoning · displayed cost $14.95/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 62% |
| 10 | GPT-6 Astralower-cost setting· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 61.1% |
| 11 | GPT-6 AstraTerminal-Bench Science 0.1 · OpenAI release chart · xhigh reasoning · displayed cost $15.76/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 60.9% |
| 12 | Claude Sonnet 5.5· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 59.9% |
| 13 | Claude Sonnet 5.5adaptive thinking · max effort · default sampling · 5-trial mean · context ≤1M unless noted · Claude Code --bare · max · no internet · safeguards with no fallback activation· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 59.9% |
| 14 | Claude Opus 5.5· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 58.7% |
| 15 | Claude Opus 5.5Opus 5.5 System Card capability section · Claude Code --bare · max · safeguards with fallback · 10 trials/task· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 58.7% |
| 16 | GPT-6.1 Terra· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 58.3% |
| 17 | Gemini 4 Argon· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 57.6% |
| 18 | GPT-6 AstraTerminal-Bench Science 0.1 · OpenAI release chart · medium reasoning · displayed cost $12.34/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 57.4% |
| 19 | GPT-6.1 Sol· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 57% |
| 20 | GPT-6.1 Soladaptive thinking · max effort · default sampling · 5-trial mean · context ≤1M unless noted · OpenAI-reported · max· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 57% |
| 21 | GPT-6.1 SolTerminal-Bench Science 0.1 · OpenAI release chart · max reasoning · displayed cost $5.47/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 57% |
| 22 | GPT-6 AstraTerminal-Bench Science 0.1 · OpenAI release chart · low reasoning · displayed cost $11.41/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 55.4% |
| 23 | GPT-6.1 SolTerminal-Bench Science 0.1 · OpenAI release chart · xhigh reasoning · displayed cost $2.89/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 53.7% |
| 24 | Claude Fable 5.1· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 52.6% |
| 25 | Claude Fable 5.1Claude Code harness· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 52.6% |
| 26 | Claude Fable 5.1OpenAI launch table· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 52.6% |
| 27 | Claude Fable 5.1Opus 5.5 System Card capability section · Claude Code --bare · max · 10 trials/task· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 52.6% |
| 28 | Claude Fable 5.1System Card capability summary · adaptive thinking · max effort · default sampling · 5 trials · context ≤1M unless noted· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 52.6% |
| 29 | GPT-6.1 SolTerminal-Bench Science 0.1 · OpenAI release chart · high reasoning · displayed cost $2.76/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 51.1% |
| 30 | GPT-6.1 SolTerminal-Bench Science 0.1 · OpenAI release chart · medium reasoning · displayed cost $2.34/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 47.6% |
| 31 | GPT-6.1 SolTerminal-Bench Science 0.1 · OpenAI release chart · low reasoning · displayed cost $1.79/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 43.7% |
| 32 | Kimi K3.1 Pro· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 38.1% |
| 33 | Claude Opus 5OpenAI launch table· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 30% |
| 34 | Claude Opus 5Claude Code harness· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 29% |
| 35 | Claude Opus 5Opus 5.5 System Card capability section · Claude Code --bare · max · 12 trials/task· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 29% |
| 36 | Claude Opus 5System Card capability summary · adaptive thinking · max effort · default sampling · 5 trials · context ≤1M unless noted· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 29% |
| 37 | GPT-6 Soladaptive thinking · max effort · default sampling · 5-trial mean · context ≤1M unless noted · OpenAI-reported · max· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 27.6% |
| 38 | GPT-6 SolTerminal-Bench Science 0.1 · OpenAI release chart · max reasoning · displayed cost $12.18/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 27.6% |
| 39 | GPT-6 SolTerminal-Bench Science 0.1 · OpenAI release chart · xhigh reasoning · displayed cost $6.77/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 25.3% |
| 40 | Claude Fable 5Claude Code harness· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 24.7% |
| 41 | Claude Fable 5Opus 5.5 System Card capability section · Claude Code --bare · max · 10 trials/task· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 24.7% |
| 42 | Claude Fable 5System Card capability summary · adaptive thinking · max effort · default sampling · 5 trials · context ≤1M unless noted· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 24.7% |
| 43 | GPT-5.6 SolClaude Code harness· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 22.4% |
| 44 | GPT-5.6 SolOpenAI launch table· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 22.4% |
| 45 | GPT-5.6 SolSystem Card capability summary · adaptive thinking · max effort · default sampling · 5 trials · context ≤1M unless noted· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 22.4% |
| 46 | Claude Fable 5OpenAI launch table· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 21.4% |
| 47 | ERNIE 5.5 Flash Preview· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 21% |
| 48 | Claude Haiku 5.5· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 20.6% |
| 49 | Claude Haiku 5.5adaptive thinking · max effort · default sampling · 5-trial mean · context ≤1M unless noted · Claude Code --bare · max · 10 trials/task · no internet · safeguards stop trial· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 20.6% |
| 50 | GPT-6 SolTerminal-Bench Science 0.1 · OpenAI release chart · high reasoning · displayed cost $4.63/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 14.6% |
Showing 1–50 of 52 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 |
|---|---|
| Claude Opus 5.5 with fallbacks (OpenAI chart label)Terminal-Bench Science 0.1 · OpenAI release chart · max reasoning · displayed cost $23.21/task· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 63.3% |
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