Benchmark Almanac

AEC-Bench (Nomic)

Claude Opus 5.5 70.2 leads since September 2026
Not in model cards

Agents search construction drawings and specifications to answer questions and identify conflicts. Task verifiers compare their outputs with known findings and award partial credit.

The 196 tasks span nine types, from reading one drawing sheet to reconciling documents across a project.

Made by
Nomic and coauthors
Released
March 2026
Kind
Tasks from real jobs
Level
University
Scored
Checked automatically
Answers
Agent task
Tasks
All public
Measures
Capability
Language
English
In and out
Text, Documents → Actions, Text
Interaction
Agentic
Environment
Terminal
Items from
Real work and use
Refresh
One fixed set
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 5.5 leads with 70.2

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 70.2
  2. 2 DeepSeek V4.1 Flash 1017 67.6
  3. 3 GPT-6.1 Sol medium 66.3
  4. 4 GPT-6 Astra medium 66.2
  5. 5 Gemini 4 Argon medium 65.3
  6. 6 Claude Fable 5 medium 64.6
  7. 7 DeepSeek V4.1 Flash 64.1
  8. 8 Kimi K3 63.5
  9. 9 Claude Opus 5 medium 63.3
  10. 10 GLM-5.3-Flash 61.9
  11. 11 Grok 4.7 medium 60.8
  12. 12 Qwen 3.8 27B 59.6
  13. 13 Claude Sonnet 5.5 medium 58.6
  14. 14 Muse Spark 1.3 medium 56.9
  15. 15 Gemini 3.7 Flash medium 55.4
  16. 16 Claude Sonnet 5 medium 55.3
  17. 17 Gemini 3.1 Pro medium 43.4
  18. 18 Muse Glimmer 35.7
  19. 19 Gemma 4 31B IT 11.8
  20. 20 Nemotron 3 Nano Omni 30B-A3B 6.20

Source: ONESTRUCTION’s leaderboard

#ModelScoreSettingDateLinkSource
1 Claude Opus 5.5 70.2 medium, pi Sep 2026 ai.onestruction.com Official leaderboard
2 DeepSeek V4.1 Flash 1017 67.6 pi Oct 2026 ai.onestruction.com Official leaderboard
3 GPT-6.1 Sol 66.3 medium, pi Sep 2026 ai.onestruction.com Official leaderboard
4 GPT-6 Astra 66.2 medium, pi Sep 2026 ai.onestruction.com Official leaderboard
5 Gemini 4 Argon 65.3 medium, pi Sep 2026 ai.onestruction.com Official leaderboard
6 Claude Fable 5 64.6 medium, pi Jun 2026 ai.onestruction.com Official leaderboard
7 DeepSeek V4.1 Flash 64.1 pi Sep 2026 ai.onestruction.com Official leaderboard
8 Kimi K3 63.5 pi Jul 2026 ai.onestruction.com Official leaderboard
9 Claude Opus 5 63.3 medium, pi Jul 2026 ai.onestruction.com Official leaderboard
10 GLM-5.3-Flash 61.9 pi Aug 2026 ai.onestruction.com Official leaderboard
11 Grok 4.7 60.8 medium, pi Sep 2026 ai.onestruction.com Official leaderboard
12 Qwen 3.8 27B 59.6 pi Aug 2026 ai.onestruction.com Official leaderboard
13 Claude Sonnet 5.5 58.6 medium, pi Sep 2026 ai.onestruction.com Official leaderboard
14 Muse Spark 1.3 56.9 medium, pi Sep 2026 ai.onestruction.com Official leaderboard
15 Gemini 3.7 Flash 55.4 medium, pi Aug 2026 ai.onestruction.com Official leaderboard
16 Claude Sonnet 5 55.3 medium, pi Jun 2026 ai.onestruction.com Official leaderboard
17 Gemini 3.1 Pro 43.4 medium, pi Feb 2026 ai.onestruction.com Official leaderboard
18 Muse Glimmer 35.7 pi Aug 2026 ai.onestruction.com Official leaderboard
19 Gemma 4 31B IT 11.8 pi Apr 2026 ai.onestruction.com Official leaderboard
20 Nemotron 3 Nano Omni 30B-A3B 6.20 pi Apr 2026 ai.onestruction.com Official leaderboard

The best score, 70.2, costs $0.33 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 OpenAI Cost frontier One model at each effort level
0.00 20.0 40.0 60.0 80.0 $0.10 $0.20 $0.50 $1.00 $2.00 $5.00 $10 Cost per task, log scale Claude Opus 5.5, medium GPT-6.1 Sol, medium 0.00 20.0 40.0 60.0 80.0 $0.10 $1.00 $10 Cost per task, log scale Claude Opus 5.5, medium

Source: ONESTRUCTION’s leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Claude Opus 5.5 medium pi 70.2 $0.33
DeepSeek V4.1 Flash 1017 pi 67.6
GPT-6.1 Sol medium pi 66.3 $0.12
GPT-6 Astra medium pi 66.2 $0.78
Gemini 4 Argon medium pi 65.3 $0.40
Claude Fable 5 medium pi 64.6 $1.60
DeepSeek V4.1 Flash pi 64.1
Kimi K3 pi 63.5
Claude Opus 5 medium pi 63.3 $1.80
GLM-5.3-Flash pi 61.9
Grok 4.7 medium pi 60.8 $0.94
Qwen 3.8 27B pi 59.6
Claude Sonnet 5.5 medium pi 58.6 $0.14
Muse Spark 1.3 medium pi 56.9 $0.62
Gemini 3.7 Flash medium pi 55.4 $0.28
Claude Sonnet 5 medium pi 55.3 $0.84
Gemini 3.1 Pro medium pi 43.4 $0.56
Muse Glimmer pi 35.7
Gemma 4 31B IT pi 11.8
Nemotron 3 Nano Omni 30B-A3B pi 6.20

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

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

Google DeepMind Anthropic OpenAI
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
0.00 20.0 40.0 60.0 80.0 2026 0.00 20.0 40.0 60.0 80.0 2026

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 Economic work

All 210 in Economic work
  • JobBench In use 130 tasks from 35 occupations workers most want delegated. Agent works through messy reference files, graded by fact-anchored binary rubrics. GLM-5.3 61.4%
  • GDPval In use Real work tasks from 44 occupations with reference files. Experts blindly compare model deliverables with expert work. Metric is win-or-tie rate. Seed2.1 Pro 87.9%
  • GDPval-AA v2.1 In use Agents with shell and web access complete 220 GDPval gold tasks from 44 occupations. Blind pairwise comparisons give Elo ratings. Claude Opus 5.5 1,846
  • AA-Briefcase v1.1 In use Version 1.1 of Artificial Analysis's multi-hour office work evaluation, with Elo ratings fitted by Crowd-BT. Claude Opus 5.5 1,822
  • Agents' Last Exam v1.1 In use Agents complete 122 public long-horizon professional workflow tasks across 55 sub-industries. Deterministic graders check outcomes against hidden references. GPT-6.2 Astra 47.5%
  • CoWorkBench In use Agent completes long-horizon productivity tasks across computer science, finance, law and medicine. Qwen in-house scoring (method unpublished). Claude Fable 5 75.9%

Data audit

All 10 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 4 pass 1 note
  • 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).
  • Note Headroom Scored as a plain number, with no fixed ceiling.
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