OmniMech
Models write CAD programs to reconstruct parts from dimensioned drawings. The evaluator compares the reconstructed geometry with reference parts and measures relative volume error.
Nine general-purpose models have results on 5,000 one-shot cases and 2,000 agent cases. The board initially sorts by relative volume error.
- Made by
- Pennsylvania State University and collaborators
- Released
- August 2026
- Kind
- CAD models
- Level
- University
- Scored
- Checked automatically
- Answers
- Code
- Measures
- Capability
- Language
- English
- Interaction
- Agentic
- Environment
- APIs and tools
- Items from
- Real work and use
- Refresh
- One fixed set
- Run by
- A host
- Strongest model tested
- GPT-5.5 #21 on the capability index
- Last model added
- June 2026
- Scores refreshed
- November 20th, 2026
GPT-5.5 leads with 15.9% among models released by June 2026 GPT-5.5 leads with 15.9% among models released by June 2026
Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. This view shows only models we track, each at its best.
The newest model with a score here came out in June 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6.2 Astra
- Claude Fable 5.2
-
Gemini 4 Carbon
- GPT-6 Astra
- and 16 more
- 1 GPT-5.5 Agent / Fixed 8 renders 15.9%
- 2
Gemini 3.5 Flash Zero-shot / Draw. + 8 renders 17.7%
- 3
Gemini 3.1 Pro Zero-shot / Draw. + 8 renders 19%
- 4 Claude Opus 4.8 Agent / Fixed 8 renders 31.2%
- 5
Qwen3.7-Plus Agent / Fixed 8 renders 54.2%
- 6 GPT-5.4 Mini Agent / Fixed 8 renders 95.7%
- 7 Ministral 14B 2512 Zero-shot / 2D Drawing 102.5%
- 8 HY3D-2 Zero-shot / Image-to-3D 109.4%
- 9
Qwen3.5-9B Agent / Adaptive renders 138.2%
- 10
Llama 4 Maverick Zero-shot / Draw. + 20 renders 184.4%
- 11 SAM 3D Zero-shot / Image-to-3D 218.3%
- 12 TRELLIS Zero-shot / Image-to-3D 317.9%
- 13 Cad Coder Zero-shot / Image-to-CAD 1395.1%
- 14 InstantMesh Zero-shot / Image-to-3D 5515.5%
Source: Official OmniMech leaderboard
- 1 GPT-5.5 Agent / Fixed 8 renders 15.9%
- 2
Gemini 3.5 Flash Zero-shot / Draw. + 8 renders 17.7%
- 3
Gemini 3.1 Pro Zero-shot / Draw. + 8 renders 19%
- 4 Claude Opus 4.8 Agent / Fixed 8 renders 31.2%
- 5
Qwen3.7-Plus Agent / Fixed 8 renders 54.2%
- 6 GPT-5.4 Mini Agent / Fixed 8 renders 95.7%
- 7
Qwen3.5-9B Agent / Adaptive renders 138.2%
- 8
Llama 4 Maverick Zero-shot / Draw. + 20 renders 184.4%
Source: Official OmniMech leaderboard
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | GPT-5.5 | 15.9% | Agent / Fixed 8 renders | Apr 2026 | Official leaderboard |
| 2 | | 17.7% | Zero-shot / Draw. + 8 renders | May 2026 | Official leaderboard |
| 3 | | 19% | Zero-shot / Draw. + 8 renders | Feb 2026 | Official leaderboard |
| 4 | Claude Opus 4.8 | 31.2% | Agent / Fixed 8 renders | May 2026 | Official leaderboard |
| 5 | | 54.2% | Agent / Fixed 8 renders | Jun 2026 | Official leaderboard |
| 6 | GPT-5.4 Mini | 95.7% | Agent / Fixed 8 renders | Mar 2026 | Official leaderboard |
| 7 | Ministral 14B 2512 | 102.5% | Zero-shot / 2D Drawing | Official leaderboard | |
| 8 | HY3D-2 | 109.4% | Zero-shot / Image-to-3D | Official leaderboard | |
| 9 | | 138.2% | Agent / Adaptive renders | Feb 2026 | Official leaderboard |
| 10 | | 184.4% | Zero-shot / Draw. + 20 renders | Apr 2025 | Official leaderboard |
| 11 | SAM 3D | 218.3% | Zero-shot / Image-to-3D | Official leaderboard | |
| 12 | TRELLIS | 317.9% | Zero-shot / Image-to-3D | Official leaderboard | |
| 13 | Cad Coder | 1395.1% | Zero-shot / Image-to-CAD | Official leaderboard | |
| 14 | InstantMesh | 5515.5% | Zero-shot / Image-to-3D | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | GPT-5.5 | 15.9% | Agent / Fixed 8 renders | Apr 2026 | Official leaderboard |
| 2 | | 17.7% | Zero-shot / Draw. + 8 renders | May 2026 | Official leaderboard |
| 3 | | 19% | Zero-shot / Draw. + 8 renders | Feb 2026 | Official leaderboard |
| 4 | Claude Opus 4.8 | 31.2% | Agent / Fixed 8 renders | May 2026 | Official leaderboard |
| 5 | | 54.2% | Agent / Fixed 8 renders | Jun 2026 | Official leaderboard |
| 6 | GPT-5.4 Mini | 95.7% | Agent / Fixed 8 renders | Mar 2026 | Official leaderboard |
| 7 | | 138.2% | Agent / Adaptive renders | Feb 2026 | Official leaderboard |
| 8 | | 184.4% | Zero-shot / Draw. + 20 renders | Apr 2025 | Official leaderboard |
Scores at each effort level
Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.
Source: Official OmniMech leaderboard
| Model | Effort | Harness | Score | Cost per task | Output tokens |
|---|---|---|---|---|---|
| Zero-shot / 2D Drawing | 448.1% | 19,773 | |||
| Zero-shot / 2D Drawing | 283.4% | 757 | |||
| Zero-shot / Draw. + 20 renders | 277.4% | 13,342 | |||
| Agent / Fixed 8 renders | 269.4% | 70,450 | |||
| Zero-shot / Draw. + 8 renders | 186.8% | 10,427 | |||
| Zero-shot / Draw. + 20 renders | 184.4% | 751 | |||
| Zero-shot / Draw. + 8 renders | 153.1% | 19,533 | |||
| GPT-5.4 Mini | Zero-shot / Draw. + 20 renders | 145.8% | 1,122 | ||
| Agent / Adaptive renders | 138.2% | 57,711 | |||
| GPT-5.4 Mini | Agent / Adaptive renders | 131.8% | 8,203 | ||
| GPT-5.4 Mini | Zero-shot / 2D Drawing | 126.7% | 1,192 | ||
| GPT-5.4 Mini | Zero-shot / Draw. + 8 renders | 97% | 1,134 | ||
| GPT-5.4 Mini | Agent / Fixed 8 renders | 95.7% | 7,998 | ||
| Zero-shot / Draw. + 20 renders | 79.4% | 17,223 | |||
| Agent / Adaptive renders | 58.2% | 48,527 | |||
| Zero-shot / 2D Drawing | 56% | 12,203 | |||
| Zero-shot / Draw. + 8 renders | 55% | 15,658 | |||
| Agent / Fixed 8 renders | 54.2% | 49,112 | |||
| GPT-5.5 | Zero-shot / 2D Drawing | 41.6% | 14,864 | ||
| Agent / Adaptive renders | 39.9% | 36,338 | |||
| GPT-5.5 | Agent / Adaptive renders | 37.7% | 15,842 | ||
| Claude Opus 4.8 | Zero-shot / 2D Drawing | 36.9% | 2,774 | ||
| Claude Opus 4.8 | Zero-shot / Draw. + 8 renders | 36.2% | 2,935 | ||
| Agent / Adaptive renders | 35.6% | 49,609 | |||
| Claude Opus 4.8 | Agent / Adaptive renders | 32.8% | 8,420 | ||
| Claude Opus 4.8 | Zero-shot / Draw. + 20 renders | 31.8% | 2,925 | ||
| GPT-5.5 | Zero-shot / Draw. + 20 renders | 31.5% | 15,186 | ||
| Claude Opus 4.8 | Agent / Fixed 8 renders | 31.2% | 9,550 | ||
| Agent / Fixed 8 renders | 30.3% | 45,364 | |||
| Agent / Fixed 8 renders | 30.3% | 32,689 | |||
| GPT-5.5 | Zero-shot / Draw. + 8 renders | 25.8% | 15,687 | ||
| Zero-shot / 2D Drawing | 23.7% | 22,037 | |||
| Zero-shot / Draw. + video | 22% | 21,917 | |||
| Zero-shot / Draw. + video | 20.9% | 14,771 | |||
| Zero-shot / Draw. + 8 renders | 19% | 22,628 | |||
| Zero-shot / 2D Drawing | 18.1% | 13,541 | |||
| Zero-shot / Draw. + 8 renders | 17.7% | 14,465 | |||
| GPT-5.5 | Agent / Fixed 8 renders | 15.9% | 12,850 |
Scores at each effort level
Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.
Source: Official OmniMech leaderboard
| Model | Effort | Harness | Score | Cost per task | Output tokens |
|---|---|---|---|---|---|
| Zero-shot / 2D Drawing | 448.1% | 19,773 | |||
| Zero-shot / 2D Drawing | 283.4% | 757 | |||
| Zero-shot / Draw. + 20 renders | 277.4% | 13,342 | |||
| Agent / Fixed 8 renders | 269.4% | 70,450 | |||
| Zero-shot / Draw. + 8 renders | 186.8% | 10,427 | |||
| Zero-shot / Draw. + 20 renders | 184.4% | 751 | |||
| Zero-shot / Draw. + 8 renders | 153.1% | 19,533 | |||
| GPT-5.4 Mini | Zero-shot / Draw. + 20 renders | 145.8% | 1,122 | ||
| Agent / Adaptive renders | 138.2% | 57,711 | |||
| GPT-5.4 Mini | Agent / Adaptive renders | 131.8% | 8,203 | ||
| GPT-5.4 Mini | Zero-shot / 2D Drawing | 126.7% | 1,192 | ||
| GPT-5.4 Mini | Zero-shot / Draw. + 8 renders | 97% | 1,134 | ||
| GPT-5.4 Mini | Agent / Fixed 8 renders | 95.7% | 7,998 | ||
| Zero-shot / Draw. + 20 renders | 79.4% | 17,223 | |||
| Agent / Adaptive renders | 58.2% | 48,527 | |||
| Zero-shot / 2D Drawing | 56% | 12,203 | |||
| Zero-shot / Draw. + 8 renders | 55% | 15,658 | |||
| Agent / Fixed 8 renders | 54.2% | 49,112 | |||
| GPT-5.5 | Zero-shot / 2D Drawing | 41.6% | 14,864 | ||
| Agent / Adaptive renders | 39.9% | 36,338 | |||
| GPT-5.5 | Agent / Adaptive renders | 37.7% | 15,842 | ||
| Claude Opus 4.8 | Zero-shot / 2D Drawing | 36.9% | 2,774 | ||
| Claude Opus 4.8 | Zero-shot / Draw. + 8 renders | 36.2% | 2,935 | ||
| Agent / Adaptive renders | 35.6% | 49,609 | |||
| Claude Opus 4.8 | Agent / Adaptive renders | 32.8% | 8,420 | ||
| Claude Opus 4.8 | Zero-shot / Draw. + 20 renders | 31.8% | 2,925 | ||
| GPT-5.5 | Zero-shot / Draw. + 20 renders | 31.5% | 15,186 | ||
| Claude Opus 4.8 | Agent / Fixed 8 renders | 31.2% | 9,550 | ||
| Agent / Fixed 8 renders | 30.3% | 45,364 | |||
| Agent / Fixed 8 renders | 30.3% | 32,689 | |||
| GPT-5.5 | Zero-shot / Draw. + 8 renders | 25.8% | 15,687 | ||
| Zero-shot / 2D Drawing | 23.7% | 22,037 | |||
| Zero-shot / Draw. + video | 22% | 21,917 | |||
| Zero-shot / Draw. + video | 20.9% | 14,771 | |||
| Zero-shot / Draw. + 8 renders | 19% | 22,628 | |||
| Zero-shot / 2D Drawing | 18.1% | 13,541 | |||
| Zero-shot / Draw. + 8 renders | 17.7% | 14,465 | |||
| GPT-5.5 | Agent / Fixed 8 renders | 15.9% | 12,850 |
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official OmniMech leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official OmniMech leaderboard
No lab model card reports it yet
We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.
More in Engineering and CAD
- EEBench In use 13 electrical-engineering circuit and PCB design tasks graded by deterministic build checks, simulation and BOM constraints. No LLM judge.
Gemini 4 Carbon 83.9%
- 3DCodeBench In use Agentic procedural 3D modeling via code: author engine-ready 3D assets through software APIs. Graded for executability and shape fidelity.
LongCat-2.5 55.4%
- CADGenBench Only in system cards Agents construct CAD models from design prompts. The generation split is graded for executability and geometric correctness. Claude Fable 5 45.1%
- PartBench Only in system cards Parametric mechanical part design and CAD-style generation, graded for geometric and manufacturing fidelity. Grok 4.6 58.5%
- Razavi-Bench Not in model cards Models explain what happens in analog circuits and why. A language model awards up to four points per answer against a reference and a rubric.
Gemini 4 Carbon 98.3%
- CircuitRubric Not in model cards Models write analog circuit netlists from design requests. A program compares transistor connections and relative device sizes with accepted reference circuits. Claude Opus 4.8 96%
Data audit
8 of 9 checks pass, 1 warning. Look first at the Task access check. We don’t know whether the tasks are public. Every benchmark
Catalog record 1 warning 1 pass
- Warning Task access We don’t know whether the tasks are public.
- Pass Confidence Checked against a primary source.
Scores 3 pass 2 notes
- Pass Independent runs Scores from the official leaderboard (14 entries).
- Pass Models scored 8 models we track, plus 6 other entries on its leaderboard.
- Pass Newest score The newest is from June 2026, for Qwen3.7-Plus.
- Note Frontier models None of the top 10 has a score. The strongest tested is GPT-5.5 (#21).
- Note Headroom Lower is better, so there’s no ceiling to reach.
Leaderboard import 3 pass
- Pass Importer We read the board on omnieda.com: 38 rows for models we track and 6 other entries at the last read.
- Pass Last read November 20th, 2026.
- Pass Read worked The last read worked.
Consistency 1 pass 1 note
- Note Lab claims No model has both a lab number and an independent run.
- Pass Known issues No published audit found problems.