Benchmark Almanac

Prompt-to-Bench-16

Gemma 4 26B A4B 10.0 leads since October 2026
Not in model cards

Models write OpenSCAD programs for laboratory fixtures from measured dimensions. They receive feedback after each failed attempt. The grader checks the rendered geometry.

The author reports 44 complete runs for 16 models on the same desktop computer. Each run covers all 16 tasks, with up to three attempts each.

Released
October 2026
Kind
CAD models
Level
University
Scored
Checked automatically
Answers
Agent task
Tasks
All public
Measures
Capability
Language
English
In and out
Text → Code
Interaction
Multi-turn
Refresh
One fixed set
Run by
Anyone
In views
CAD
Strongest model tested
Qwen 3.8 27B #59 on the capability index
Last model added
August 2026
Scores refreshed
November 20th, 2026
16
Entries on its own leaderboard, 12 models we track

Gemma 4 26B A4B leads with 10.0 among models released by August 2026 Gemma 4 26B A4B leads with 10.0 among models released by August 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 August 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 Gemma 4 26B A4B Thinking on, 16,384 tokens 10.0
  2. 2 gpt-oss-120b Thinking on, 16,384 tokens 8.50
  3. 3 Qwen 3.6 35B-A3B Thinking on, 16,384 tokens 5.50
  4. 4 Qwen3.6 27B Thinking off, 2,048 tokens 5.30
  5. 5 Ornith-1.0 35B (MoE) Thinking on, 16,384 tokens 4.00
  6. 6 gpt-oss-20b Thinking on, 16,384 tokens 4.00
  7. 7 Qwen 3.8 27B Thinking off, 2,048 tokens 4.00
  8. 8 Qwen3-Coder-Next 80B-A3B Thinking off, 2,048 tokens 3.00
  9. 9 Qwen3-30B-A3B-Thinking Jul 2025, Thinking on, 16,384 tokens 2.00
  10. 10 Qwen3.5-35B-A3B Thinking on, 16,384 tokens 2.00
  11. 11 GLM-4.7-Flash 30B-A3B Thinking on, 16,384 tokens 1.00
  12. 12 Qwen3.5 9B (Tiiny) Thinking off, 2,048 tokens 1.00
  13. 13 Qwen3-8B Thinking on, 16,384 tokens 1.00
  14. 14 Qwen3-30B-A3B-Instruct Jul 2025, Thinking off, 2,048 tokens 0.70
  15. 15 Qwen3-Coder-30B-A3B-Instruct Thinking off, 2,048 tokens 0.00
  16. 16 qwen3-4b-instruct-2507 Thinking off, 2,048 tokens 0.00

Source: Official Prompt-to-Bench-16 leaderboard

#EntryScoreSettingDateSource
1 Gemma 4 26B A4B 10.0 Thinking on, 16,384 tokens Apr 2026 Official leaderboard
2 gpt-oss-120b 8.50 Thinking on, 16,384 tokens Aug 2025 Official leaderboard
3 Qwen 3.6 35B-A3B 5.50 Thinking on, 16,384 tokens Apr 2026 Official leaderboard
4 Qwen3.6 27B 5.30 Thinking off, 2,048 tokens Apr 2026 Official leaderboard
5 Ornith-1.0 35B (MoE) 4.00 Thinking on, 16,384 tokens Official leaderboard
6 gpt-oss-20b 4.00 Thinking on, 16,384 tokens Aug 2025 Official leaderboard
7 Qwen 3.8 27B 4.00 Thinking off, 2,048 tokens Aug 2026 Official leaderboard
8 Qwen3-Coder-Next 80B-A3B 3.00 Thinking off, 2,048 tokens Official leaderboard
9 Qwen3-30B-A3B-Thinking Jul 2025 2.00 Thinking on, 16,384 tokens Jul 2025 Official leaderboard
10 Qwen3.5-35B-A3B 2.00 Thinking on, 16,384 tokens Feb 2026 Official leaderboard
11 GLM-4.7-Flash 30B-A3B 1.00 Thinking on, 16,384 tokens Official leaderboard
12 Qwen3.5 9B (Tiiny) 1.00 Thinking off, 2,048 tokens Official leaderboard
13 Qwen3-8B 1.00 Thinking on, 16,384 tokens Apr 2025 Official leaderboard
14 Qwen3-30B-A3B-Instruct Jul 2025 0.70 Thinking off, 2,048 tokens Jul 2025 Official leaderboard
15 Qwen3-Coder-30B-A3B-Instruct 0.00 Thinking off, 2,048 tokens Jul 2025 Official leaderboard
16 qwen3-4b-instruct-2507 0.00 Thinking off, 2,048 tokens Aug 2025 Official leaderboard
#ModelScoreSettingDateSource
1 Gemma 4 26B A4B 10.0 Thinking on, 16,384 tokens Apr 2026 Official leaderboard
2 gpt-oss-120b 8.50 Thinking on, 16,384 tokens Aug 2025 Official leaderboard
3 Qwen 3.6 35B-A3B 5.50 Thinking on, 16,384 tokens Apr 2026 Official leaderboard
4 Qwen3.6 27B 5.30 Thinking off, 2,048 tokens Apr 2026 Official leaderboard
5 gpt-oss-20b 4.00 Thinking on, 16,384 tokens Aug 2025 Official leaderboard
6 Qwen 3.8 27B 4.00 Thinking off, 2,048 tokens Aug 2026 Official leaderboard
7 Qwen3-30B-A3B-Thinking Jul 2025 2.00 Thinking on, 16,384 tokens Jul 2025 Official leaderboard
8 Qwen3.5-35B-A3B 2.00 Thinking on, 16,384 tokens Feb 2026 Official leaderboard
9 Qwen3-8B 1.00 Thinking on, 16,384 tokens Apr 2025 Official leaderboard
10 Qwen3-30B-A3B-Instruct Jul 2025 0.70 Thinking off, 2,048 tokens Jul 2025 Official leaderboard
11 Qwen3-Coder-30B-A3B-Instruct 0.00 Thinking off, 2,048 tokens Jul 2025 Official leaderboard
12 qwen3-4b-instruct-2507 0.00 Thinking off, 2,048 tokens Aug 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.

Lowest effort to highest, light to dark
Gemma 4 26B A4B
Thinking of…
10.0
gpt-oss-120b
Thinking of… Thinking on…
8.50
Qwen 3.6 35B-A3B
Thinking of… Thinking on…
5.50
gpt-oss-20b
Thinking of… Thinking on…
4.00
Qwen3-30B-A3B-Thinking
Thinking of… Thinking on…
2.00
Qwen3.5-35B-A3B
Thinking of… Thinking on…
2.00
Qwen3-8B
Thinking of… Thinking on…
1.00

Source: Official Prompt-to-Bench-16 leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Gemma 4 26B A4B Thinking on, 16,384 tokens 10.0
Gemma 4 26B A4B Thinking off, 2,048 tokens 9.30
gpt-oss-120b Thinking on, 16,384 tokens 8.50
Qwen 3.6 35B-A3B Thinking on, 16,384 tokens 5.50
gpt-oss-120b Thinking off, 2,048 tokens 5.30
Qwen3.6 27B Thinking off, 2,048 tokens 5.30
gpt-oss-20b Thinking on, 16,384 tokens 4.00
Qwen 3.8 27B Thinking off, 2,048 tokens 4.00
Qwen3-30B-A3B-Thinking Jul 2025 Thinking on, 16,384 tokens 2.00
Qwen3.5-35B-A3B Thinking on, 16,384 tokens 2.00
Qwen 3.6 35B-A3B Thinking off, 2,048 tokens 2.00
Qwen3-8B Thinking on, 16,384 tokens 1.00
Qwen3.5-35B-A3B Thinking off, 2,048 tokens 1.00
Qwen3-30B-A3B-Instruct Jul 2025 Thinking off, 2,048 tokens 0.70
Qwen3-8B Thinking off, 2,048 tokens 0.00
Qwen3-30B-A3B-Thinking Jul 2025 Thinking off, 2,048 tokens 0.00
Qwen3-Coder-30B-A3B-Instruct Thinking off, 2,048 tokens 0.00
gpt-oss-20b Thinking off, 2,048 tokens 0.00
qwen3-4b-instruct-2507 Thinking off, 2,048 tokens 0.00

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.

Lowest effort to highest, light to dark
Gemma 4 26B A4B
Thinking of…
10.0
gpt-oss-120b
Thinking of… Thinking on…
8.50
Qwen 3.6 35B-A3B
Thinking of… Thinking on…
5.50
gpt-oss-20b
Thinking of… Thinking on…
4.00
Qwen3-30B-A3B-Thinking
Thinking of… Thinking on…
2.00
Qwen3.5-35B-A3B
Thinking of… Thinking on…
2.00
Qwen3-8B
Thinking of… Thinking on…
1.00

Source: Official Prompt-to-Bench-16 leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Gemma 4 26B A4B Thinking on, 16,384 tokens 10.0
Gemma 4 26B A4B Thinking off, 2,048 tokens 9.30
gpt-oss-120b Thinking on, 16,384 tokens 8.50
Qwen 3.6 35B-A3B Thinking on, 16,384 tokens 5.50
gpt-oss-120b Thinking off, 2,048 tokens 5.30
Qwen3.6 27B Thinking off, 2,048 tokens 5.30
gpt-oss-20b Thinking on, 16,384 tokens 4.00
Qwen 3.8 27B Thinking off, 2,048 tokens 4.00
Qwen3-30B-A3B-Thinking Jul 2025 Thinking on, 16,384 tokens 2.00
Qwen3.5-35B-A3B Thinking on, 16,384 tokens 2.00
Qwen 3.6 35B-A3B Thinking off, 2,048 tokens 2.00
Qwen3-8B Thinking on, 16,384 tokens 1.00
Qwen3.5-35B-A3B Thinking off, 2,048 tokens 1.00
Qwen3-30B-A3B-Instruct Jul 2025 Thinking off, 2,048 tokens 0.70
Qwen3-8B Thinking off, 2,048 tokens 0.00
Qwen3-30B-A3B-Thinking Jul 2025 Thinking off, 2,048 tokens 0.00
Qwen3-Coder-30B-A3B-Instruct Thinking off, 2,048 tokens 0.00
gpt-oss-20b Thinking off, 2,048 tokens 0.00
qwen3-4b-instruct-2507 Thinking off, 2,048 tokens 0.00

Number one over time

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

Google DeepMind
Own leaderboard Not tested since Oct 2026

Source: Official Prompt-to-Bench-16 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
0.00 2.00 4.00 6.00 8.00 10.0 2025 2026 0.00 2.00 4.00 6.00 8.00 10.0 2025 2026

Source: Official Prompt-to-Bench-16 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

All 9 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 3 pass 2 notes
  • Pass Independent runs Scores from the official leaderboard (16 entries).
  • Pass Models scored 12 models we track, plus 4 other entries on its leaderboard.
  • Pass Newest score The newest is from August 2026, for Qwen 3.8 27B.
  • Note Frontier models None of the top 10 has a score. The strongest tested is Qwen 3.8 27B (#59).
  • Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on github.com: 19 rows for models we track and 4 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