Benchmark Almanac

IFC-Bench v2

Claude Opus 5 65.8% leads since July 2026
Not in model cards

Agents inspect IFC files and answer questions about building elements and their relationships. A model grades each answer against a reference.

The original release had 1,027 questions across 37 files. The authors removed a duplicate training question and retained all 514 test questions.

Released
February 2026
Kind
CAD models
Level
University
Scored
Graded by a model
Answers
Agent task
Tasks
All public
Measures
Capability
Language
English
In and out
Text, 3D → Actions, Text
Interaction
Agentic
Environment
APIs and tools
Items from
Real work and use
Run by
Anyone
In views
Agents
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 leads with 65.8%

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 medium 65.8%
  2. 2 Claude Opus 5.5 medium 60.5%
  3. 3 Gemini 4 Argon medium 59.1%
  4. 4 Kimi K3 58.6%
  5. 5 Muse Spark 1.3 medium 57.8%
  6. 6 Claude Fable 5 medium 54.7%
  7. 7 Claude Sonnet 5.5 medium 54.7%
  8. 8 Grok 4.7 medium 54.3%
  9. 9 GLM-5.3-Flash 51.8%
  10. 10 Gemini 3.7 Flash medium 51.6%
  11. 11 DeepSeek V4.1 Flash max 51%
  12. 12 Qwen 3.8 27B 49.2%
  13. 13 GPT-6.1 Sol medium 47.7%
  14. 14 GPT-6 Astra medium 46.7%
  15. 15 Claude Sonnet 5 medium 46.5%
  16. 16 DeepSeek V4.1 Flash 1017 max 45.1%
  17. 17 Muse Glimmer 44.9%
  18. 18 Gemini 3.1 Pro medium 40.5%
  19. 19 Gemma 4 31B IT 26.3%
  20. 20 Nemotron 3 Nano Omni 30B-A3B 3.11%

Source: ONESTRUCTION’s leaderboard

#ModelScoreSettingReleasedRun onSource
1 Claude Opus 5 65.8% medium Jul 2026 Oct 2026 Official leaderboard
2 Claude Opus 5.5 60.5% medium Sep 2026 Oct 2026 Official leaderboard
3 Gemini 4 Argon 59.1% medium Sep 2026 Oct 2026 Official leaderboard
4 Kimi K3 58.6% Jul 2026 Oct 2026 Official leaderboard
5 Muse Spark 1.3 57.8% medium Sep 2026 Oct 2026 Official leaderboard
6 Claude Fable 5 54.7% medium Jun 2026 Oct 2026 Official leaderboard
7 Claude Sonnet 5.5 54.7% medium Sep 2026 Oct 2026 Official leaderboard
8 Grok 4.7 54.3% medium Sep 2026 Oct 2026 Official leaderboard
9 GLM-5.3-Flash 51.8% Aug 2026 Oct 2026 Official leaderboard
10 Gemini 3.7 Flash 51.6% medium Aug 2026 Oct 2026 Official leaderboard
11 DeepSeek V4.1 Flash 51% max Sep 2026 Oct 2026 Official leaderboard
12 Qwen 3.8 27B 49.2% Aug 2026 Oct 2026 Official leaderboard
13 GPT-6.1 Sol 47.7% medium Sep 2026 Oct 2026 Official leaderboard
14 GPT-6 Astra 46.7% medium Sep 2026 Oct 2026 Official leaderboard
15 Claude Sonnet 5 46.5% medium Jun 2026 Oct 2026 Official leaderboard
16 DeepSeek V4.1 Flash 1017 45.1% max Oct 2026 Oct 2026 Official leaderboard
17 Muse Glimmer 44.9% Aug 2026 Oct 2026 Official leaderboard
18 Gemini 3.1 Pro 40.5% medium Feb 2026 Oct 2026 Official leaderboard
19 Gemma 4 31B IT 26.3% Apr 2026 Oct 2026 Official leaderboard
20 Nemotron 3 Nano Omni 30B-A3B 3.11% Apr 2026 Oct 2026 Official leaderboard

The best score, 65.8%, costs $0.17 a task

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.

Anthropic Google DeepMind Cost frontier One model at each effort level
0% 20% 40% 60% 80% $0.01 $0.02 $0.05 $0.10 $0.20 $0.50 $1.00 Cost per task, log scale Claude Opus 5, medium Gemini 4 Argon, medium Claude Sonnet 5.5, medium 0% 20% 40% 60% 80% $0.01 $0.10 $1.00 Cost per task, log scale Claude Opus 5, medium Gemini 4 Argon, medium

Source: ONESTRUCTION’s leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Claude Opus 5 medium 65.8% $0.17 3,367
Claude Opus 5.5 medium 60.5% $0.31 10,719
Gemini 4 Argon medium 59.1% $0.06 3,004
Kimi K3 58.6% 6,476
Muse Spark 1.3 medium 57.8% $0.08 6,494
Claude Fable 5 medium 54.7% $0.23 1,935
Claude Sonnet 5.5 medium 54.7% $0.05 2,589
Grok 4.7 medium 54.3% $0.14 7,079
GLM-5.3-Flash 51.8% 31,930
Gemini 3.7 Flash medium 51.6% $0.02 3,074
DeepSeek V4.1 Flash max 51% 103,377
Qwen 3.8 27B 49.2% 75,773
GPT-6.1 Sol medium 47.7% $0.03 1,160
GPT-6 Astra medium 46.7% $0.13 1,226
Claude Sonnet 5 medium 46.5% $0.07 3,735
DeepSeek V4.1 Flash 1017 max 45.1% 27,669
Muse Glimmer 44.9% 21,428
Gemini 3.1 Pro medium 40.5% $0.05 3,061
Gemma 4 31B IT 26.3% 10,469
Nemotron 3 Nano Omni 30B-A3B 3.11% 37,863

4 models have held number one on ONESTRUCTION’s leaderboard since its release in February 2026

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

Google DeepMind Anthropic Moonshot AI
ONESTRUCTION leaderboard 4 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 Kimi K3 Gemini 3.1 Pro 0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) Claude Opus 5 Kimi K3

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 65.8%, 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