Benchmark Almanac

Ishigaki-IDS-Bench

Claude Opus 5.5 63% leads since September 2026
Not in model cards

Models convert building information requirements into IDS XML. An audit tool checks validity, and a scorer compares each requirement with the reference file.

The authors translated 83 scenarios into Japanese and English for 166 examples. Some tasks ask the model to revise an earlier specification.

Made by
ONESTRUCTION
Released
May 2026
Kind
CAD models
Level
University
Scored
Checked automatically
Answers
Code
Tasks
All public
Measures
Capability
Language
Japanese, English
In and out
Text, Code → Code
Interaction
Single turn
Items from
Translated
Refresh
One fixed set
Run by
Anyone
Strongest model tested
GPT-6 Astra #4 on the capability index
Last model added
October 2026
Scores refreshed
November 20th, 2026
20
Models scored on ONESTRUCTION’s leaderboard

Claude Opus 5.5 leads with 63%

Best score per model from ONESTRUCTION’s 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.

  1. 1 Claude Opus 5.5 medium 63%
  2. 2 Gemini 4 Argon medium 61.9%
  3. 3 GPT-6 Astra medium 61.7%
  4. 4 Claude Opus 5 medium 60.9%
  5. 5 GPT-6.1 Sol medium 60.7%
  6. 6 Claude Sonnet 5.5 medium 60%
  7. 7 Claude Fable 5 medium 58.6%
  8. 8 Grok 4.7 medium 58.2%
  9. 9 Gemini 3.7 Flash medium 52.5%
  10. 10 Gemini 3.1 Pro medium 50.1%
  11. 11 Muse Spark 1.3 medium 47.4%
  12. 12 Kimi K3 46.5%
  13. 13 Claude Sonnet 5 medium 45%
  14. 14 DeepSeek V4.1 Flash 1017 44.5%
  15. 15 DeepSeek V4.1 Flash 42.5%
  16. 16 GLM-5.3-Flash 38.2%
  17. 17 Gemma 4 31B IT 24.3%
  18. 18 Muse Glimmer 22.9%
  19. 19 Qwen 3.8 27B 19.8%
  20. 20 Nemotron 3 Nano Omni 30B-A3B 17.8%

Source: ONESTRUCTION’s leaderboard

#ModelScoreSettingDateLinkSource
1 Claude Opus 5.5 63% medium, pi Sep 2026 ai.onestruction.com Official leaderboard
2 Gemini 4 Argon 61.9% medium, pi Sep 2026 ai.onestruction.com Official leaderboard
3 GPT-6 Astra 61.7% medium, pi Sep 2026 ai.onestruction.com Official leaderboard
4 Claude Opus 5 60.9% medium, pi Jul 2026 ai.onestruction.com Official leaderboard
5 GPT-6.1 Sol 60.7% medium, pi Sep 2026 ai.onestruction.com Official leaderboard
6 Claude Sonnet 5.5 60% medium, pi Sep 2026 ai.onestruction.com Official leaderboard
7 Claude Fable 5 58.6% medium, pi Jun 2026 ai.onestruction.com Official leaderboard
8 Grok 4.7 58.2% medium, pi Sep 2026 ai.onestruction.com Official leaderboard
9 Gemini 3.7 Flash 52.5% medium, pi Aug 2026 ai.onestruction.com Official leaderboard
10 Gemini 3.1 Pro 50.1% medium, pi Feb 2026 ai.onestruction.com Official leaderboard
11 Muse Spark 1.3 47.4% medium, pi Sep 2026 ai.onestruction.com Official leaderboard
12 Kimi K3 46.5% pi Jul 2026 ai.onestruction.com Official leaderboard
13 Claude Sonnet 5 45% medium, pi Jun 2026 ai.onestruction.com Official leaderboard
14 DeepSeek V4.1 Flash 1017 44.5% pi Oct 2026 ai.onestruction.com Official leaderboard
15 DeepSeek V4.1 Flash 42.5% pi Sep 2026 ai.onestruction.com Official leaderboard
16 GLM-5.3-Flash 38.2% pi Aug 2026 ai.onestruction.com Official leaderboard
17 Gemma 4 31B IT 24.3% pi Apr 2026 ai.onestruction.com Official leaderboard
18 Muse Glimmer 22.9% pi Aug 2026 ai.onestruction.com Official leaderboard
19 Qwen 3.8 27B 19.8% pi Aug 2026 ai.onestruction.com Official leaderboard
20 Nemotron 3 Nano Omni 30B-A3B 17.8% pi Apr 2026 ai.onestruction.com Official leaderboard

The best score, 63%, costs $0.04 a task

Runs on the cost frontier score higher than any run that's cheaper. GPT-6.1 Sol at medium comes within 2.3 points of the best for $0.01 a task, 4 times cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

Google DeepMind Anthropic Cost frontier One model at each effort level
0% 20% 40% 60% 80% $0.001 $0.002 $0.005 $0.01 $0.02 $0.05 $0.10 Cost per task, log scale Claude Opus 5.5, medium Gemini 4 Argon, medium Gemini 3.7 Flash, medium 0% 20% 40% 60% 80% $0.001 $0.01 $0.10 Cost per task, log scale Claude Opus 5.5, medium Gemini 3.7 Flash, medium

Source: ONESTRUCTION’s leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Claude Opus 5.5 medium pi 63% $0.04
Gemini 4 Argon medium pi 61.9% $0.01
GPT-6 Astra medium pi 61.7% $0.07
Claude Opus 5 medium pi 60.9% $0.05
GPT-6.1 Sol medium pi 60.7% $0.01
Claude Sonnet 5.5 medium pi 60% $0.02
Claude Fable 5 medium pi 58.6% $0.09
Grok 4.7 medium pi 58.2% $0.03
Gemini 3.7 Flash medium pi 52.5% $0.009
Gemini 3.1 Pro medium pi 50.1% $0.03
Muse Spark 1.3 medium pi 47.4% $0.02
Kimi K3 pi 46.5%
Claude Sonnet 5 medium pi 45% $0.02
DeepSeek V4.1 Flash 1017 pi 44.5%
DeepSeek V4.1 Flash pi 42.5%
GLM-5.3-Flash pi 38.2%
Gemma 4 31B IT pi 24.3%
Muse Glimmer pi 22.9%
Qwen 3.8 27B pi 19.8%
Nemotron 3 Nano Omni 30B-A3B pi 17.8%

5 models have held number one on ONESTRUCTION’s leaderboard since its release in May 2026

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

Google DeepMind Anthropic OpenAI
ONESTRUCTION leaderboard 5 leaders

Source: ONESTRUCTION’s 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 ONESTRUCTION’s leaderboard Best so far
0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) Claude Opus 5.5 Claude Fable 5 Gemini 3.1 Pro 0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) Claude Opus 5.5 Gemini 3.1 Pro

Source: ONESTRUCTION’s 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 11 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 5 pass
  • Pass Independent runs Scores from the official leaderboard (20 entries).
  • Pass Models scored 20 models we track.
  • Pass Newest score The newest is from October 2026, for DeepSeek V4.1 Flash 1017.
  • Pass Frontier models 5 of the top 10 have a score, led by GPT-6 Astra (#4).
  • Pass Headroom The best is 63%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on ONESTRUCTION: 20 rows for models we track 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