Benchmark Almanac

cad-spec

DeepSeek-V3 Mar 2025 78% leads since August 2026
Not in model cards

Models write CadQuery code for mounting plates from dimensioned specifications. The evaluator measures each finished solid against nine requirements.

On November 19, 2026, DeepSeek Chat V3 0324 led 16 models at 78% across four held-out tiers. GPT-4.1 nano scored 63%, and Qwen3 Coder Next scored 61%.

Made by
cad-spec team
Released
August 2026
Kind
CAD models
Level
University
Scored
Checked automatically
Answers
Code
Tasks
All public
Measures
Capability
Language
English
In and out
Text → Code
Interaction
Single turn
Refresh
One fixed set
Run by
A host
In views
CAD
Strongest model tested
Qwen3-235B-A22B-Instruct Jul 2025 #133 on the capability index
Last model added
February 2026
Scores refreshed
November 20th, 2026
16
Entries on its own leaderboard, 14 models we track

DeepSeek-V3 leads with 78% among models released by February 2026 DeepSeek-V3 leads with 78% among models released by February 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 February 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 DeepSeek-V3 Mar 2025, greedy 78%
  2. 2 GPT-4.1 nano greedy 63%
  3. 3 Qwen3-Coder-Next greedy 61%
  4. 4 Qwen3-235B-A22B-Instruct Jul 2025, greedy 59%
  5. 5 GPT-4o mini greedy 40%
  6. 6 Llama 4 Maverick greedy 33%
  7. 7 Mistral Small 3.2 greedy 32%
  8. 8 Qwen3-Coder-30B-A3B-Instruct greedy 28%
  9. 9 Phi-4 greedy 18%
  10. 10 Qwen3-30B-A3B-Instruct Jul 2025, greedy 18%
  11. 11 Llama 3.3 70B Instruct greedy 16%
  12. 12 mistralai/codestral-2508 greedy 13%
  13. 13 Gemma 3 27B greedy 12%
  14. 14 meta-llama/llama-3.1-8b-instruct greedy 8.00%
  15. 15 Gemma 3 4B greedy 8.00%
  16. 16 Qwen2.5-Coder (1.5B) greedy 7.00%

Source: Official cad-spec leaderboard

#EntryScore ±SettingDateSource
1 DeepSeek-V3 Mar 2025 78% 72 to 85 greedy Mar 2025 Official leaderboard
2 GPT-4.1 nano 63% 56 to 71 greedy Apr 2025 Official leaderboard
3 Qwen3-Coder-Next 61% ±7 greedy Feb 2026 Official leaderboard
4 Qwen3-235B-A22B-Instruct Jul 2025 59% ±8 greedy Jul 2025 Official leaderboard
5 GPT-4o mini 40% 33 to 46 greedy Jul 2024 Official leaderboard
6 Llama 4 Maverick 33% 26 to 41 greedy Apr 2025 Official leaderboard
7 Mistral Small 3.2 32% 25 to 40 greedy Jun 2025 Official leaderboard
8 Qwen3-Coder-30B-A3B-Instruct 28% 22 to 36 greedy Jul 2025 Official leaderboard
9 Phi-4 18% ±5 greedy Dec 2024 Official leaderboard
10 Qwen3-30B-A3B-Instruct Jul 2025 18% ±4 greedy Jul 2025 Official leaderboard
11 Llama 3.3 70B Instruct 16% ±4 greedy Dec 2024 Official leaderboard
12 mistralai/codestral-2508 13% 9 to 18 greedy Official leaderboard
13 Gemma 3 27B 12% ±4 greedy Mar 2025 Official leaderboard
14 meta-llama/llama-3.1-8b-instruct 8.00% ±4 greedy Official leaderboard
15 Gemma 3 4B 8.00% ±4 greedy Mar 2025 Official leaderboard
16 Qwen2.5-Coder (1.5B) 7.00% ±4 greedy Sep 2024 Official leaderboard
#ModelScore ±SettingDateSource
1 DeepSeek-V3 Mar 2025 78% 72 to 85 greedy Mar 2025 Official leaderboard
2 GPT-4.1 nano 63% 56 to 71 greedy Apr 2025 Official leaderboard
3 Qwen3-Coder-Next 61% ±7 greedy Feb 2026 Official leaderboard
4 Qwen3-235B-A22B-Instruct Jul 2025 59% ±8 greedy Jul 2025 Official leaderboard
5 GPT-4o mini 40% 33 to 46 greedy Jul 2024 Official leaderboard
6 Llama 4 Maverick 33% 26 to 41 greedy Apr 2025 Official leaderboard
7 Mistral Small 3.2 32% 25 to 40 greedy Jun 2025 Official leaderboard
8 Qwen3-Coder-30B-A3B-Instruct 28% 22 to 36 greedy Jul 2025 Official leaderboard
9 Phi-4 18% ±5 greedy Dec 2024 Official leaderboard
10 Qwen3-30B-A3B-Instruct Jul 2025 18% ±4 greedy Jul 2025 Official leaderboard
11 Llama 3.3 70B Instruct 16% ±4 greedy Dec 2024 Official leaderboard
12 Gemma 3 27B 12% ±4 greedy Mar 2025 Official leaderboard
13 Gemma 3 4B 8.00% ±4 greedy Mar 2025 Official leaderboard
14 Qwen2.5-Coder (1.5B) 7.00% ±4 greedy Sep 2024 Official leaderboard

Rankings by part

We read 5 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as L0, L1 and L2. GPT-4o mini leads 1 of them, and 4 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.

L0

14 models we track have a score here.

  1. 1 GPT-4o mini ↑4 100%
  2. 2 Qwen2.5-Coder (1.5B) ↑12 100%
  3. 3 Llama 3.3 70B Instruct ↑8 100%
  4. 4 Gemma 3 27B ↑8 100%
  5. 5 DeepSeek-V3 Mar 2025 ↓4 100%
  6. 6 Qwen3-235B-A22B-Instruct Jul 2025 ↓2 100%
  7. 7 Gemma 3 4B ↑6 97%
  8. 8 GPT-4.1 nano ↓6 97%
  9. 9 Qwen3-Coder-Next ↓6 97%
  10. 10 Phi-4 ↓1 93%

The best score, 78%, costs $0.03 for the full run

Runs on the cost frontier score higher than any run that's cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

Microsoft Google DeepMind DeepSeek OpenAI Mistral AI Cost frontier One model at each effort level
0% 20% 40% 60% 80% 100% $0.001 $0.002 $0.005 $0.01 $0.02 $0.05 $0.10 Cost of the full run, log scale DeepSeek-V3, greedy GPT-4.1 nano, greedy Mistral Small 3.2, greedy Phi-4, greedy Gemma 3 4B, greedy 0% 20% 40% 60% 80% 100% $0.001 $0.01 $0.10 Cost of the full run, log scale DeepSeek-V3, greedy GPT-4.1 nano, greedy

Source: Official cad-spec leaderboard

ModelEffortHarnessScoreCost of the full runOutput tokens
DeepSeek-V3 Mar 2025 greedy 78% $0.03
GPT-4.1 nano greedy 63% $0.01
Qwen3-Coder-Next greedy 61% $0.04
Qwen3-235B-A22B-Instruct Jul 2025 greedy 59% $0.02
GPT-4o mini greedy 40% $0.02
Llama 4 Maverick greedy 33% $0.03
Mistral Small 3.2 greedy 32% $0.008
Qwen3-Coder-30B-A3B-Instruct greedy 28% $0.01
Phi-4 greedy 18% $0.007
Qwen3-30B-A3B-Instruct Jul 2025 greedy 18% $0.01
Llama 3.3 70B Instruct greedy 16% $0.02
Gemma 3 27B greedy 12% $0.01
Gemma 3 4B greedy 8.00% $0.005
Qwen2.5-Coder (1.5B) greedy 7.00%

The best score, 78%, costs $0.03 for the full run

Runs on the cost frontier score higher than any run that's cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

Microsoft Google DeepMind DeepSeek OpenAI Mistral AI Cost frontier One model at each effort level
0% 20% 40% 60% 80% 100% $0.001 $0.002 $0.005 $0.01 $0.02 $0.05 $0.10 Cost of the full run, log scale DeepSeek-V3, greedy GPT-4.1 nano, greedy Mistral Small 3.2, greedy Phi-4, greedy Gemma 3 4B, greedy 0% 20% 40% 60% 80% 100% $0.001 $0.01 $0.10 Cost of the full run, log scale DeepSeek-V3, greedy GPT-4.1 nano, greedy

Source: Official cad-spec leaderboard

ModelEffortHarnessScoreCost of the full runOutput tokens
DeepSeek-V3 Mar 2025 greedy 78% $0.03
GPT-4.1 nano greedy 63% $0.01
Qwen3-Coder-Next greedy 61% $0.04
Qwen3-235B-A22B-Instruct Jul 2025 greedy 59% $0.02
GPT-4o mini greedy 40% $0.02
Llama 4 Maverick greedy 33% $0.03
Mistral Small 3.2 greedy 32% $0.008
Qwen3-Coder-30B-A3B-Instruct greedy 28% $0.01
Phi-4 greedy 18% $0.007
Qwen3-30B-A3B-Instruct Jul 2025 greedy 18% $0.01
Llama 3.3 70B Instruct greedy 16% $0.02
Gemma 3 27B greedy 12% $0.01
Gemma 3 4B greedy 8.00% $0.005
Qwen2.5-Coder (1.5B) greedy 7.00%

Number one over time

Lab claims count from the day the card came out, independent runs from the model’s release.

DeepSeek
Own leaderboard Not tested since Aug 2026

Source: Official cad-spec leaderboard

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

Plot with other benchmarks
On the benchmark’s own leaderboard Best so far
0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) DeepSeek-V3 GPT-4o mini 0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) DeepSeek-V3 GPT-4o mini

Source: Official cad-spec 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

All 112 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

9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 9 months ago, in February 2026, for Qwen3-Coder-Next. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
  • Pass Independent runs Scores from the official leaderboard (16 entries).
  • Pass Models scored 14 models we track, plus 2 other entries on its leaderboard.
  • Warning Newest score The newest came 9 months ago, in February 2026, for Qwen3-Coder-Next.
  • Note Frontier models None of the top 10 has a score. The strongest tested is Qwen3-235B-A22B-Instruct (#133).
  • Pass Headroom The best is 78%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on github.com: 14 rows for models we track and 2 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.

How we check