Domains: ScienceCoding
Skills: CodingAgenticTool UseReasoningInstruction Following

Four-Factor Integrity Breakdown & Primary-Source Evidence

Privacy

Weight 40%
0/10
Access Model: Public Plaintext
"Dataset: https://github.com/harbor-framework/terminal-bench-science Implementation: We evaluate the full Terminal-Bench-Science 0.1.0 release (70 tasks: 19 life sciences, 17 physical sciences, 17 mathematical sciences, 9 engineering sciences, 8 earth sciences) using the mini-swe-agent harness, with pass@1 scoring averaged over 3 repeats per task"

Originality

Weight 25%
10/10
Reuse Profile: All Original (0% reuse)
"Terminal-Bench-Science is an open academic collaboration developed by Stanford University researchers with the Terminal-Bench and Harbor team. It includes contributions from scientists at institutions across the globe, comprising 70 expert-curated and reviewed research workflow tasks drawn from real scientific practice."

Freshness

Weight 20%
6/10
Initial Release: 2026-08Latest Refresh: 2026-08Rolling Updates: No
"Written by Steven Dillmann on August 27, 2026 Terminal-Bench-Science is a benchmark led by researchers at Stanford University and built by the team behind Terminal-Bench in collaboration with domain experts from a range of scientific disciplines and research institutions around the world."

Grading

Weight 15%
3.3/10
Integrity: Self-Reported / Live NetworkSandbox: live internet
"network_mode = "public" # anti-cheat rests on the scrubbed lake identity, not on network isolation"
Primary source (GitHub) ↗ Verified 2026-11-17

Frontier Model Standings on Terminal-Bench-Science 0.1

Top 15 active models tracked
Rank Model Official Score Percentile
#1GPT-6.2 Astra max OpenAI78.1%100.0p
#2Claude Fable 5.2 max Anthropic73.3%100.0p
#3Gemini 4 Neon xhigh Google70.0%100.0p
#4GPT-6.2 Sol max OpenAI67.6%97.5p
#5Gemini 4 Carbon high Google64.8%95.0p
#6GPT-6 Astra max OpenAI63.3%92.4p
#7Claude Opus 5.5 max Anthropic59.1%89.9p
#8GPT-6.1 Sol max OpenAI58.1%87.4p
#9GPT-6.1 Terra max OpenAI57.6%84.9p
#10Claude Sonnet 5.5 max Anthropic53.3%82.4p
#11Gemini 4 Argon high Google52.9%79.9p
#12Gemini 4 Hydrogen xhigh Google49.5%77.4p
#13Claude Fable 5.1 max Anthropic43.3%74.9p
#14Muse Wonder Full max Meta37.1%72.3p
#15Kimi K3.1 Pro max Kimi34.3%69.8p

Embeddable Audit Badge

View Markdown (.md)

Benchmark maintainers can embed this live SVG badge in GitHub READMEs or documentation:

BenchAudit Terminal-Bench-Science 0.1 Tier C 4.2/10
[![BenchAudit Tier C](https://benchaudit.org/badges/terminal-bench-science-0-1.svg)](https://benchaudit.org/benchmarks/terminal-bench-science-0-1/)