CADGenBench Generation

We show this table for reference; we do not rank on it.

Hugging Face's open benchmark for generating valid and geometrically correct 3D mechanical parts.

Generation score on CADGenBench Generation — September 23, 2026 snapshot

We mirror the published generation score view for CADGenBench Generation. Godela leads the public snapshot at 84.5%, followed by build123d-mcp-v0381-claude-opus-5-xhigh-full-r6 (59.6%) and gpt-5.6-sol-xhigh-floor-chain-v0379 (45.1%). We do not use these results to rank models overall.

Godela

84.5%

build123d-mcp-v0381-claude-opus-5-xhigh-full-r6

Anthropic

59.6%

gpt-5.6-sol-xhigh-floor-chain-v0379

OpenAI

45.1%

24 modelsCodingCurrentDisplay onlyUpdated September 23, 2026 snapshot

Generation score table (24 models)

Score

GodelaGodela

84.5%

build123d-mcp-v0381-claude-opus-5-xhigh-full-r6Anthropic

59.6%

gpt-5.6-sol-xhigh-floor-chain-v0379OpenAI

45.1%

Archie in ForgeSatvik Adyanthaya

42.3%

Grok 4.6 high Mecado Baseline with Build123dxAI

41.5%

Grok 4.6 xhigh Mecado Baseline with Build123dxAI

39.1%

Claude Fable 5Anthropic · Closed

37.3%

gpt-5.5-build123d-mcp-0.3.59-xhigh-r5OpenAI

37.2%

GPT-5.6 Sol xhigh Mecado Baseline with Build123dOpenAI

36.4%

Claude Opus 5 max Mecado Baseline with Build123dAnthropic

36.1%

GPT-5.6 Sol HF Baseline with Build123dOpenAI

35.5%

gpt-5.5-build123d-mcp-0.3.56-v2OpenAI

34.5%

GPT-5.5 Pro HF Baseline with Build123dOpenAI

32.1%

build123d-mcp + Claude Opus 4.8 (full, 81 fixtures)Anthropic

31.5%

Claude Opus 4.7 HF Baseline with Build123dAnthropic

29.9%

GPT-5.5OpenAI · Closed

29.7%

Opus 4.8 — Claude Code + text-to-cad skillAnthropic

28.0%

Claude Opus 4.8Anthropic · Closed

27.4%

Claude Opus 4.6 HF Baseline with Build123dAnthropic

23.6%

Gemini 3.1 Flash-Lite HF Baseline with Build123dGoogle

22.4%

Claude Sonnet 4.6Anthropic · Closed

22.1%

Gemini 3.1 ProGoogle · Closed

21.1%

GLM-4.6V HF Baseline with Build123dZ.AI

16.4%

Qwen3-VL 235B-A22B Instruct HF Baseline with Build123dAlibaba

16.3%

How we show CADGenBench Generation

We mirror 24 completed submissions that the official CADGenBench dataset marks validated. The visible score is the generation-task score; aggregate, editing, validity, version, and data-revision fields remain attached to each row.

CADGenBench Generation is display only. Each submission can include a different agent, CAD library, prompt, and tool configuration, so we do not treat the table as a normalized base-model comparison.

Snapshot

24 validated submissions49 generation tasks5 mapped baselinesDisplay only

Validated submissions dataset CADGenBench GitHub repository

The published CADGenBench Generation snapshot places Godela first at 84.5%. The third row is 39.4 points behind. The broader top-10 range is 48.4 points, so the table still separates the published systems.

24 models have been evaluated on CADGenBench Generation. The benchmark falls in the Coding category. CADGenBench Generation is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.

About CADGenBench Generation

Year

2026

Tasks

49 CAD generation samples

Format

Validity-gated CAD score

Difficulty

Professional mechanical CAD

The benchmark repository, submission dataset, and hosted leaderboard are public. BenchLM stores the generation score separately from editing and keeps the model-card rows display-only.

CADGenBench Public benchmark source

Freshness and provenance

Version

CADGenBench Generation 2026

Refresh cadence

Quarterly

Staleness state

Current

Question availability

Public benchmark set

CurrentDisplay only

BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.

Questions

What does CADGenBench Generation measure?

Hugging Face's open benchmark for generating valid and geometrically correct 3D mechanical parts.

Which model leads the published CADGenBench Generation snapshot?

Godela currently leads the published CADGenBench Generation snapshot with 84.5% generation score. BenchLM shows this benchmark for display only and does not use it in overall rankings.

How many models are evaluated on CADGenBench Generation?

The September 23, 2026 snapshot snapshot contains 24 AI models.

Last updated: September 23, 2026 snapshot · mirrored from the public benchmark leaderboard

Know when it’s worth switching models

The model to choose, the cheaper alternative, and the release we would wait on.

Read a sample issue

Join 2,000+ readers.

One email each week. Unsubscribe anytime.