Benchmark Almanac

MOMENT Eval

Claude Opus 4.7 95.6% leads since August 2026
Not in model cards

Models use a calculator and Python to solve 91 structural engineering problems. They return a value or classification that a program checks against the answer key.

CORE studio publishes 68 model configurations across six providers. The board includes tool calls and cost. Some models appear at several reasoning settings.

Released
August 2026
Kind
Tasks from real jobs
Level
Expert
Scored
Checked automatically
Answers
Short answer
Measures
Capability
Language
English
In and out
Text → Actions, Text
Interaction
Agentic
Refresh
One fixed set
Run by
A host
In views
Agents
Strongest model tested
Claude Opus 5 #11 on the capability index
Last model added
July 2026
Scores refreshed
November 20th, 2026
41
Models scored on its own leaderboard

3 models share the top score of 95.6% among models released by July 2026

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.

The newest model with a score here came out in July 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 Claude Opus 4.7 95.6%
  2. 2 Claude Fable 5 95.6%
  3. 3 GPT-5.6 Sol xhigh 95.6%
  4. 4 Claude Opus 4.5 94.5%
  5. 5 Claude Opus 4.6 94.5%
  6. 6 Gemini 3.1 Pro 94.5%
  7. 7 GPT-5.5 medium 94.5%
  8. 8 Claude Opus 5 94.5%
  9. 9 Gemini 3 Pro 93.4%
  10. 10 Claude Opus 4.8 92.3%
  11. 11 Claude Sonnet 4.6 90.1%
  12. 12 Kimi K2.6 89%
  13. 13 GPT-5.3 Codex high 87.9%
  14. 14 GPT-5.6 Terra xhigh 87.9%
  15. 15 GPT-5.4 high 86.8%
  16. 16 GPT-5.2 xhigh 85.7%
  17. 17 GPT-5 high 84.6%
  18. 18 Claude Sonnet 4.5 84.6%
  19. 19 GPT-5.6 Luna medium 84.6%
  20. 20 Kimi K2.5 83.5%
  21. 21 Grok 4.20 82.4%
  22. 22 o3 81.3%
  23. 23 Grok 4.1 Fast reasoning 81.3%
  24. 24 Gemini 3.1 Flash-Lite 80.2%
  25. 25 GPT-5 mini medium 76.9%
  26. 26 GPT-5.1 high 75.8%
  27. 27 Qwen3.5-35B-A3B 75.8%
  28. 28 o1 74.7%
  29. 29 o3-mini 73.6%
  30. 30 o4-mini 73.6%
  31. 31 Claude Haiku 4.5 71.4%
  32. 32 Qwen 3.6 35B-A3B 67%
  33. 33 gpt-oss-20b 44%
  34. 34 GPT-4.1 42.9%
  35. 35 GPT-4o Mar 2025 38.5%
  36. 36 Gemma 4 31B IT 38.5%
  37. 37 Qwen2.5-32B 31.9%
  38. 38 Qwen3-32B 26.4%
  39. 39 Gemma 4 E4B 16.5%
  40. 40 Gemma 4 26B A4B 12.1%
  41. 41 Gemma 4 E2B 2.20%

Source: Official MOMENT Eval leaderboard

#ModelScoreSettingDateSource
1 Claude Opus 4.7 95.6% Apr 2026 Official leaderboard
2 Claude Fable 5 95.6% Jun 2026 Official leaderboard
3 GPT-5.6 Sol 95.6% xhigh Jul 2026 Official leaderboard
4 Claude Opus 4.5 94.5% Nov 2025 Official leaderboard
5 Claude Opus 4.6 94.5% Feb 2026 Official leaderboard
6 Gemini 3.1 Pro 94.5% Feb 2026 Official leaderboard
7 GPT-5.5 94.5% medium Apr 2026 Official leaderboard
8 Claude Opus 5 94.5% Jul 2026 Official leaderboard
9 Gemini 3 Pro 93.4% Nov 2025 Official leaderboard
10 Claude Opus 4.8 92.3% May 2026 Official leaderboard
11 Claude Sonnet 4.6 90.1% Feb 2026 Official leaderboard
12 Kimi K2.6 89% Apr 2026 Official leaderboard
13 GPT-5.3 Codex 87.9% high Feb 2026 Official leaderboard
14 GPT-5.6 Terra 87.9% xhigh Jul 2026 Official leaderboard
15 GPT-5.4 86.8% high Mar 2026 Official leaderboard
16 GPT-5.2 85.7% xhigh Dec 2025 Official leaderboard
17 GPT-5 84.6% high Aug 2025 Official leaderboard
18 Claude Sonnet 4.5 84.6% Sep 2025 Official leaderboard
19 GPT-5.6 Luna 84.6% medium Jul 2026 Official leaderboard
20 Kimi K2.5 83.5% Jan 2026 Official leaderboard
21 Grok 4.20 82.4% Feb 2026 Official leaderboard
22 o3 81.3% Apr 2025 Official leaderboard
23 Grok 4.1 Fast 81.3% reasoning Nov 2025 Official leaderboard
24 Gemini 3.1 Flash-Lite 80.2% Mar 2026 Official leaderboard
25 GPT-5 mini 76.9% medium Aug 2025 Official leaderboard
26 GPT-5.1 75.8% high Nov 2025 Official leaderboard
27 Qwen3.5-35B-A3B 75.8% Feb 2026 Official leaderboard
28 o1 74.7% Dec 2024 Official leaderboard
29 o3-mini 73.6% Jan 2025 Official leaderboard
30 o4-mini 73.6% Apr 2025 Official leaderboard
31 Claude Haiku 4.5 71.4% Oct 2025 Official leaderboard
32 Qwen 3.6 35B-A3B 67% Apr 2026 Official leaderboard
33 gpt-oss-20b 44% Aug 2025 Official leaderboard
34 GPT-4.1 42.9% Apr 2025 Official leaderboard
35 GPT-4o Mar 2025 38.5% Mar 2025 Official leaderboard
36 Gemma 4 31B IT 38.5% Apr 2026 Official leaderboard
37 Qwen2.5-32B 31.9% Sep 2024 Official leaderboard
38 Qwen3-32B 26.4% Apr 2025 Official leaderboard
39 Gemma 4 E4B 16.5% Mar 2026 Official leaderboard
40 Gemma 4 26B A4B 12.1% Apr 2026 Official leaderboard
41 Gemma 4 E2B 2.20% Mar 2026 Official leaderboard

The best score, 95.6%, costs $3.58 for the full run

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.

xAI OpenAI Cost frontier One model at each effort level
0% 20% 40% 60% 80% 100% $0.10 $1.00 $10 $100 Cost of the full run, log scale GPT-5.6 Sol, xhigh GPT-5.6 Sol, low GPT-5.6 Luna, medium Grok 4.1 Fast, non-reasoning 0% 20% 40% 60% 80% 100% $0.10 $1.00 $10 $100 Cost of the full run, log scale GPT-5.6 Sol, xhigh GPT-5.6 Sol, low

Source: Official MOMENT Eval leaderboard

ModelEffortHarnessScoreCost of the full runOutput tokens
Claude Opus 4.7 95.6% $6.69
Claude Fable 5 95.6% $11
GPT-5.6 Sol xhigh 95.6% $3.58
Claude Opus 4.5 94.5% $15
Claude Opus 4.6 94.5% $15
Gemini 3.1 Pro 94.5% $8.01
GPT-5.5 medium 94.5% $4.19
GPT-5.6 Sol high 94.5% $2.63
Claude Opus 5 94.5% $5.42
Gemini 3 Pro 93.4% $6.71
GPT-5.5 high 93.4% $7.38
GPT-5.5 xhigh 93.4% $8.54
GPT-5.6 Sol medium 93.4% $2.35
Claude Opus 4.8 92.3% $3.68
Claude Sonnet 4.6 90.1% $7.29
GPT-5.5 low 90.1% $2.45
GPT-5.6 Sol low 90.1% $1.90
Kimi K2.6 89% $5.66
GPT-5.3 Codex high 87.9% $3.53
GPT-5.6 Terra xhigh 87.9% $1.87
GPT-5.3 Codex medium 86.8% $1.74
GPT-5.4 high 86.8% $6.02
GPT-5.4 xhigh 86.8% $6.96
GPT-5.2 xhigh 85.7% $13
GPT-5 high 84.6% $10
Claude Sonnet 4.5 84.6% $15
GPT-5.6 Luna medium 84.6% $0.15
GPT-5.6 Terra medium 84.6% $1.11
Kimi K2.5 83.5% $4.72
Kimi K2.5 thinking 83.5% $5.50
GPT-5.4 low 83.5% $1.66
GPT-5.4 medium 83.5% $2.94
GPT-5.6 Luna low 83.5% $0.12
GPT-5.6 Luna high 83.5% $0.32
GPT-5.6 Terra high 83.5% $1.32
GPT-5.2 low 82.4% $1.99
GPT-5.3 Codex low 82.4% $1.23
GPT-5.3 Codex xhigh 82.4% $5.48
Grok 4.20 82.4% $1.68
GPT-5.6 Terra low 82.4% $1.04
o3 81.3% $7.14
Grok 4.1 Fast reasoning 81.3% $0.15
GPT-5.2 medium 80.2% $2.61
GPT-5.2 high 80.2% $4.91
Gemini 3.1 Flash-Lite 80.2% $0.84
GPT-5 mini medium 76.9% $0.74
GPT-5 mini high 76.9% $1.46
GPT-5.1 high 75.8% $7.51
Qwen3.5-35B-A3B 75.8%
o1 74.7% $42
o3-mini 73.6% $1.43
o4-mini 73.6% $1.50
Claude Haiku 4.5 71.4% $5.44
GPT-5 mini low 70.3% $0.33
GPT-5.6 Luna xhigh 69.2% $0.47
Grok 4.20 67% $2.03
Qwen 3.6 35B-A3B 67%
GPT-5 mini minimal 55% $0.21
Grok 4.1 Fast non-reasoning 48.4% $0.11
gpt-oss-20b 44%
GPT-4.1 42.9% $2.05
GPT-4o Mar 2025 38.5% $2.30
Gemma 4 31B IT 38.5%
Qwen2.5-32B 31.9%
Qwen3-32B 26.4%
Gemma 4 E4B 16.5%
Gemma 4 26B A4B 12.1%
Gemma 4 E2B 2.20%

Number one over time

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

Anthropic
Own leaderboard Passed 85% in Aug 2026, by Claude Opus 4.7

Source: Official MOMENT Eval leaderboard

Gemini 3 Pro on the official leaderboard passed 85% in November 2025

Models released before it came out already scored that high. 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% 2025 2026 Near the ceiling (85%) Claude Opus 4.7 Claude Opus 4.5 GPT-5 o1 Qwen2.5-32B 0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) Claude Opus 4.7 GPT-5

Source: Official MOMENT Eval 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

8 of 9 checks pass, 1 warning. Look first at the Task access check. We don’t know whether the tasks are public. Every benchmark

Catalog record 1 warning 1 pass
  • Warning Task access We don’t know whether the tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 3 pass 2 notes
  • Pass Independent runs Scores from the official leaderboard (41 entries).
  • Pass Models scored 41 models we track.
  • Pass Newest score The newest is from July 2026, for Claude Opus 5.
  • Note Frontier models None of the top 10 has a score, which is fine for a retired test. The strongest tested is Claude Opus 5 (#11).
  • Note Headroom Saturated. The best is 95.6%, past 85% since November 2025.
Leaderboard import 3 pass
  • Pass Importer We read the board on moment.corestudio.io: 68 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