LetTokenBurn
Sign in
Verified86%Harbor: pendingAgentic / Tool UseKnowledge & Reasoning

EngiWorld

EngiWorld: What Can Frontier Agents Deliver in Professional Engineering Environments?

A benchmark evaluating frontier agents on professional engineering tasks across CAD/CAE/CAM/BIM domains using GUI/CLI interfaces with artifact-centric verification.

PaperDatasetHomepage

How to run

ltb · engiworld
$ npm i -g lettokenburn-cli
$ ltb bench info engiworld

External harness. No Harbor adapter has been generated for this benchmark yet. `ltb bench info` prints the known links.

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

Scores

7 sourced entries
HarnessSourceConfidenceLink
Claude Opus 544.3N/AIndependentmediumSep 29, 2026arxiv.org
GPT-5.6 Sol38.0N/AIndependentmediumSep 29, 2026arxiv.org
Qwen3.8 Max25.8N/AIndependentmediumSep 29, 2026arxiv.org
DeepSeek V4.1 Flash25.5N/AIndependentmediumSep 29, 2026arxiv.org
Gemini 3.7 Flash25.3N/AIndependentmediumSep 29, 2026arxiv.org
Kimi K324.1N/AIndependentmediumSep 29, 2026arxiv.org
Qwen3.8 Flash16.1N/AIndependentmediumSep 29, 2026arxiv.org

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

Content & Construction: EngiWorld is the first benchmark structured around a complete design loop, featuring 1,301 expert-curated tasks spanning six engineering domains (CAD, CAE, CAM, BIM, EDA, and 3D visualization) across 26 software platforms. It covers GUI and CLI interfaces with task types ranging from software selection to open-ended challenges.

Scoring & Key Findings: The benchmark employs an artifact-centric evaluation methodology using a unified domain-verifier suite that programmatically checks geometric validity, physical feasibility, and rule compliance. Evaluation of seven frontier models reveals a substantial capability gap, where the strongest model achieves an EngiScore of only 44.3.