Benchmark Almanac

CEQuest

Gemini 3.7 Flash 95.5% leads since August 2026
Not in model cards

Models answer text questions about construction drawings, trade conventions, and quantity estimates. The original test has 101 multiple-choice questions and 63 true or false questions.

ONESTRUCTION ranks 20 models on 134 questions. It excludes 30 multiple-choice questions whose answer keys point to missing options.

Released
August 2025
Kind
Tasks from real jobs
Level
University
Scored
Checked automatically
Answers
Multiple choice
Tasks
All public
Measures
Capability
Language
English
In and out
Text → Text
Interaction
Single turn
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 its own leaderboard

Gemini 3.7 Flash leads with 95.5%

Best score per model 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.

  1. 1 Gemini 3.7 Flash medium 95.5%
  2. 2 Gemini 4 Argon medium 94%
  3. 3 Gemini 3.1 Pro medium 93.3%
  4. 4 Claude Fable 5 medium 93.3%
  5. 5 Grok 4.7 medium 93.3%
  6. 6 Claude Opus 5.5 medium 93.3%
  7. 7 GPT-6.1 Sol medium 93.3%
  8. 8 Muse Spark 1.3 medium 92.5%
  9. 9 Kimi K3 91.8%
  10. 10 Claude Opus 5 medium 91.8%
  11. 11 GPT-6 Astra medium 91.8%
  12. 12 DeepSeek V4.1 Flash 1017 91.8%
  13. 13 GLM-5.3-Flash 90.3%
  14. 14 DeepSeek V4.1 Flash 90.3%
  15. 15 Gemma 4 31B IT 89.5%
  16. 16 Qwen 3.8 27B 88.8%
  17. 17 Claude Sonnet 5 medium 87.3%
  18. 18 Claude Sonnet 5.5 medium 85.8%
  19. 19 Nemotron 3 Nano Omni 30B-A3B 85.1%
  20. 20 Muse Glimmer 85.1%

Source: Official CEQuest leaderboard

#ModelScoreSettingDateLinkSource
1 Gemini 3.7 Flash 95.5% medium, pi Aug 2026 ai.onestruction.com Official leaderboard
2 Gemini 4 Argon 94% medium, pi Sep 2026 ai.onestruction.com Official leaderboard
3 Gemini 3.1 Pro 93.3% medium, pi Feb 2026 ai.onestruction.com Official leaderboard
4 Claude Fable 5 93.3% medium, pi Jun 2026 ai.onestruction.com Official leaderboard
5 Grok 4.7 93.3% medium, pi Sep 2026 ai.onestruction.com Official leaderboard
6 Claude Opus 5.5 93.3% medium, pi Sep 2026 ai.onestruction.com Official leaderboard
7 GPT-6.1 Sol 93.3% medium, pi Sep 2026 ai.onestruction.com Official leaderboard
8 Muse Spark 1.3 92.5% medium, pi Sep 2026 ai.onestruction.com Official leaderboard
9 Kimi K3 91.8% pi Jul 2026 ai.onestruction.com Official leaderboard
10 Claude Opus 5 91.8% medium, pi Jul 2026 ai.onestruction.com Official leaderboard
11 GPT-6 Astra 91.8% medium, pi Sep 2026 ai.onestruction.com Official leaderboard
12 DeepSeek V4.1 Flash 1017 91.8% pi Oct 2026 ai.onestruction.com Official leaderboard
13 GLM-5.3-Flash 90.3% pi Aug 2026 ai.onestruction.com Official leaderboard
14 DeepSeek V4.1 Flash 90.3% pi Sep 2026 ai.onestruction.com Official leaderboard
15 Gemma 4 31B IT 89.5% pi Apr 2026 ai.onestruction.com Official leaderboard
16 Qwen 3.8 27B 88.8% pi Aug 2026 ai.onestruction.com Official leaderboard
17 Claude Sonnet 5 87.3% medium, pi Jun 2026 ai.onestruction.com Official leaderboard
18 Claude Sonnet 5.5 85.8% medium, pi Sep 2026 ai.onestruction.com Official leaderboard
19 Nemotron 3 Nano Omni 30B-A3B 85.1% pi Apr 2026 ai.onestruction.com Official leaderboard
20 Muse Glimmer 85.1% pi Aug 2026 ai.onestruction.com Official leaderboard

The best score, 95.5%, costs $0.0007 a task

Runs on the cost frontier score higher than any run that's cheaper. GPT-6.1 Sol at medium comes within 2.2 points of the best for $0.0003 a task, about half the cost. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

Anthropic Google DeepMind OpenAI Cost frontier One model at each effort level
0% 20% 40% 60% 80% 100% $0.0001 $0.0002 $0.0005 $0.001 $0.002 $0.005 $0.01 Cost per task, log scale Gemini 3.7 Flash, medium GPT-6.1 Sol, medium Claude Sonnet 5, medium 0% 20% 40% 60% 80% 100% $0.0001 $0.001 $0.01 Cost per task, log scale Gemini 3.7 Flash, medium Claude Sonnet 5, medium

Source: Official CEQuest leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Gemini 3.7 Flash medium pi 95.5% $0.0007
Gemini 4 Argon medium pi 94% $0.0006
Gemini 3.1 Pro medium pi 93.3% $0.003
Claude Fable 5 medium pi 93.3% $0.003
Grok 4.7 medium pi 93.3% $0.005
Claude Opus 5.5 medium pi 93.3% $0.002
GPT-6.1 Sol medium pi 93.3% $0.0003
Muse Spark 1.3 medium pi 92.5% $0.002
Kimi K3 pi 91.8%
Claude Opus 5 medium pi 91.8% $0.003
GPT-6 Astra medium pi 91.8% $0.001
DeepSeek V4.1 Flash 1017 pi 91.8%
GLM-5.3-Flash pi 90.3%
DeepSeek V4.1 Flash pi 90.3%
Gemma 4 31B IT pi 89.5%
Qwen 3.8 27B pi 88.8%
Claude Sonnet 5 medium pi 87.3% $0.0003
Claude Sonnet 5.5 medium pi 85.8% $0.0006
Nemotron 3 Nano Omni 30B-A3B pi 85.1%
Muse Glimmer pi 85.1%

2 models have held number one on the official leaderboard since its release in August 2025

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

Google DeepMind
Own leaderboard Passed 85% in Feb 2026, by Gemini 3.1 Pro

Source: Official CEQuest leaderboard

Gemini 3.1 Pro on the official leaderboard passed 85% in February 2026

That was 6 months after its release. By the same age, 14% of the benchmarks from 2025 had passed 85%. 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 Best so far
0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) Gemini 3.7 Flash Gemini 3.1 Pro 0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) Gemini 3.7 Flash

Source: Official CEQuest 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 Saturated. The best is 95.5%, past 85% since February 2026.
Leaderboard import 3 pass
  • Pass Importer We read the board on ai.onestruction.com: 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