LetTokenBurn
Sign in
Verified86%Harbor: pendingCodingMath

SciCode

Sci Code

SciCode: 科学计算编程

PaperRepoDatasetLeaderboardHomepage

How to run

ltb · scicode
$ npm i -g lettokenburn-cli
$ ltb bench pull scicode

External harness. pip install inspect-evals && inspect eval inspect_evals/scicode --model <provider/model> https://github.com/UKGovernmentBEIS/inspect_evals

Runs happen on your machine with your own subscription; only the score summary is pushed. CLI docs

Scores

128 sourced entries
HarnessSourceConfidenceLink
Claude Opus 5.566.9N/AIndependentmediumN/Aartificialanalysis.ai
Claude Fable 5.163.1N/AIndependentmediumN/Aartificialanalysis.ai
Gemini 4 Argon61.8N/AIndependentmediumN/Aartificialanalysis.ai
Claude Fable 561.0N/AIndependentmediumN/Aartificialanalysis.ai
Claude Sonnet 5.561.0N/AIndependentmediumN/Aartificialanalysis.ai
MiMo V2.6 Pro60.9N/AIndependentmediumN/Aartificialanalysis.ai
Gemini 3.7 Flash59.8N/AIndependentmediumN/Aartificialanalysis.ai
Muse Spark 1.359.7N/AIndependentmediumN/Aartificialanalysis.ai
Kimi K359.5N/AIndependentmediumN/Aartificialanalysis.ai
GLM-5.359.0N/AIndependentmediumN/Aartificialanalysis.ai
Gemini 3.1 Pro58.9N/AIndependentmediumN/Aartificialanalysis.ai
Step 5 Preview58.9N/AIndependentmediumN/Aartificialanalysis.ai
Muse Spark 1.158.8N/AIndependentmediumN/Aartificialanalysis.ai
GPT-6 Sol57.6N/AIndependentmediumN/Aartificialanalysis.ai
Grok-4.757.4N/AIndependentmediumN/Aartificialanalysis.ai
GPT-5.6 Sol57.1N/AIndependentmediumN/Aartificialanalysis.ai
Gemini 3.8 Flash56.6N/AIndependentmediumN/Aartificialanalysis.ai
GPT-5.456.6N/AIndependentmediumN/Aartificialanalysis.ai
GPT-6 Astra56.5N/AIndependentmediumN/Aartificialanalysis.ai
Grok-4.656.5N/AIndependentmediumN/Aartificialanalysis.ai
Claude Opus 556.4N/AIndependentmediumN/Aartificialanalysis.ai
Muse Spark 1.256.4N/AIndependentmediumN/Aartificialanalysis.ai
GPT-5.556.1N/AIndependentmediumN/Aartificialanalysis.ai
GPT-6.1 Sol55.8N/AIndependentmediumN/Aartificialanalysis.ai
GPT-5.6-terra55.0N/AIndependentmediumN/Aartificialanalysis.ai
GPT-6 Luna54.6N/AIndependentmediumN/Aartificialanalysis.ai
Claude Opus 4.754.5N/AIndependentmediumN/Aartificialanalysis.ai
Grok 4.554.1N/AIndependentmediumN/Aartificialanalysis.ai
Claude Sonnet 553.6N/AIndependentmediumN/Aartificialanalysis.ai
GPT-5.6 Luna53.6N/AIndependentmediumN/Aartificialanalysis.ai
Claude Opus 4.853.5N/AIndependentmediumN/Aartificialanalysis.ai
Kimi K2.653.5N/AIndependentmediumN/Aartificialanalysis.ai
Gemini-3.5-flash53.1N/AIndependentmediumN/Aartificialanalysis.ai
Qwen3.8 Max52.9N/AIndependentmediumN/Aartificialanalysis.ai
Gemini 3.6 Flash52.7N/AIndependentmediumN/Aartificialanalysis.ai
Qwen3.8 Max 090252.1N/AIndependentmediumN/Aartificialanalysis.ai
DeepSeek V4.1 Flash51.9N/AIndependentmediumN/Aartificialanalysis.ai
GLM-5.3 Flash51.6N/AIndependentmediumN/Aartificialanalysis.ai
Muse Spark51.5N/AIndependentmediumN/Aartificialanalysis.ai
MiMo V2.6 Flash51.3N/AIndependentmediumN/Aartificialanalysis.ai
DeepSeek V4 Pro 081351.0N/AIndependentmediumN/Aartificialanalysis.ai
GLM-5.250.5N/AIndependentmediumN/Aartificialanalysis.ai
Grok Build 0.150.2N/AIndependentmediumN/Aartificialanalysis.ai
MiMo V2.5 Pro50.2N/AIndependentmediumN/Aartificialanalysis.ai
DeepSeek V4 Pro50.0N/AIndependentmediumN/Aartificialanalysis.ai
DeepSeek V4 Flash 073149.9N/AIndependentmediumN/Aartificialanalysis.ai
GPT-5.4-mini49.9N/AIndependentmediumN/Aartificialanalysis.ai
Kimi K2.549.0N/AIndependentmediumN/Aartificialanalysis.ai
Qwen3.7 Max48.8N/AIndependentmediumN/Aartificialanalysis.ai
Inkling Small48.7N/AIndependentmediumN/Aartificialanalysis.ai
Top 50 of 128 by score

Every number links to where it was published. Low-confidence rows are muted and need per-cell verification.

Reproduced by the community

All runs
Loading community runs…

About

SciCode (Sci Code) — 科学计算编程. Scoring: Pass@1. Difficulty: 高. Imported from benchmark_research.md (2026-06).