# tbs-official-20261010.md

Source: https://www.terminal-bench-science.ai/
Retrieved: 2026-10-10

# TERMINAL-BENCH-SCIENCE [0.09876543210](https://doi.org/10.5281/zenodo.22110253)

A benchmark for evaluating AI agents on research workflows across scientific domains

[Run the benchmark](https://www.terminal-bench-science.ai/run) [View the tasks](https://hub.harborframework.com/datasets/terminal-bench-science/terminal-bench-science/latest?tab=tasks) [Contribute a task](https://www.terminal-bench-science.ai/contribute) Cite the benchmark

ALLDOMAINSLIFESCIENCESPHYSICALSCIENCESEARTHSCIENCESMATHEMATICALSCIENCESENGINEERINGSCIENCES

0.1

RANKINGPARETOMATRIXRADAR

| RANK | MODEL | AGENT | RESOLUTION RATE | RELEASE DATE | TOKENS | COST |
| --- | --- | --- | --- | --- | --- | --- |
| 1 | [GPT-6 Astra](https://developers.openai.com/api/docs/models/gpt-6-astra)(max) | [Codex](https://openai.com/codex/) | **68.1%** \+/\- 3.2% | 2026-09-03 | 2.3B | $4,997 |
| 2 | [Opus 5.5](https://www.anthropic.com/claude-opus-5-5)(max) | [Claude Code](https://claude.com/product/claude-code) | **63.3%** \+/\- 3.3% | 2026-09-22 | 8.7B | $4,874 |
| 3 | [Fable 5.1](https://www.anthropic.com/claude-fable-and-mythos-5-1)(max) | [Claude Code](https://claude.com/product/claude-code) | **40.0%** \+/\- 3.4% | 2026-09-01 | 3.8B | $6,247 |
| 4 | [Opus 5](https://www.anthropic.com/news/claude-opus-5)(max) | [Claude Code](https://claude.com/product/claude-code) | **30.0%** \+/\- 3.2% | 2026-07-24 | 7.3B | $6,993 |
| 5 | [GPT-5.6 Sol](https://developers.openai.com/api/docs/models/gpt-5.6-sol)(max) | [Codex](https://openai.com/codex/) | **22.4%** \+/\- 2.9% | 2026-07-09 | 8.4B | $4,222 |
| 6 | [Fable 5](https://www.anthropic.com/)(max) | [Claude Code](https://claude.com/product/claude-code) | **21.4%** \+/\- 2.8% | 2026-06-09 | 6.4B | $14,176 |
| 7 | [DeepSeek V4.1 Flash](https://api-docs.deepseek.com/updates/)(max) | [Codex](https://openai.com/codex/) | **15.7%** \+/\- 2.5% | 2026-09-10 | 14.9B | $386 |
| 8 | [Grok 4.7](https://x.ai/news/grok-4-7)(xhigh) | [Grok Build](https://x.ai/) | **14.3%** \+/\- 2.4% | 2026-09-21 | 4.2B | $3,071 |
| 9 | [Gemini 3.8 Flash](https://ai.google.dev/gemini-api/docs/models/gemini-3.8-flash)(high) | [mini-SWE-agent](https://mini-swe-agent.com/latest/) | **12.4%** \+/\- 2.3% | 2026-09-02 | 10.5B | $1,117 |
| 10 | [Opus 4.8](https://www.anthropic.com/)(max) | [Claude Code](https://claude.com/product/claude-code) | **10.5%** \+/\- 2.1% | 2026-05-28 | 6.8B | $5,838 |
| 11 | [GPT-5.6 Terra](https://developers.openai.com/api/docs/models/gpt-5.6-terra)(max) | [Codex](https://openai.com/codex/) | **8.6%** \+/\- 1.9% | 2026-07-09 | 7.6B | $1,984 |
| 12 | [GLM 5.3](https://z.ai/)(max) | [Claude Code](https://claude.com/product/claude-code) | **8.1%** \+/\- 1.9% | 2026-08-14 | 8.5B | $2,733 |
| 13 | [Kimi K3](https://www.moonshot.ai/)(max) | [Claude Code](https://claude.com/product/claude-code) | **7.1%** \+/\- 1.8% | 2026-07-16 | 3.2B | $1,524 |
| 14 | [Grok 4.6](https://x.ai/)(xhigh) | [Grok Build](https://x.ai/) | **7.1%** \+/\- 1.8% | 2026-08-12 | 3.7B | $3,344 |
| 15 | [Gemini 3.7 Flash](https://ai.google.dev/gemini-api/docs/models/gemini-3.7-flash)(high) | [mini-SWE-agent](https://mini-swe-agent.com/latest/) | **5.7%** \+/\- 1.6% | 2026-08-13 | 12.0B | $1,297 |
| 16 | [DeepSeek V4 Pro](https://api-docs.deepseek.com/news/news260813/)(max) | [Codex](https://openai.com/codex/) | **3.8%** \+/\- 1.3% | 2026-08-13 | 11.2B | $1,120 |
| 17 | [GPT-5.6 Luna](https://developers.openai.com/api/docs/models/gpt-5.6-luna)(max) | [Codex](https://openai.com/codex/) | **3.3%** \+/\- 1.2% | 2026-07-09 | 14.2B | $383 |

Official HTML footer: Resolution rate ranked by agent and model performance ± standard error