Benchmark Almanac

Argo-Bench

Claude Fable 5.2 70.2 leads since November 2026
Not in model cards

Agents query a simulated food delivery company's warehouse and act on their findings. The grader measures the business consequences of their decisions across 210 tasks.

Made by
TextQL
Released
October 2026
Kind
Simulated companies
Level
Expert
Scored
Checked automatically
Answers
Agent task
Tasks
Public, plus a private test set
Measures
Capability
Language
English
In and out
Text, Tables → Actions, Code
Interaction
Agentic
Items from
Generated
Refresh
One fixed set
Run by
A host
Strongest model tested
GPT-6.2 Astra #1 on the capability index
Last model added
November 2026
Scores refreshed
November 20th, 2026
24
Models scored on its own leaderboard

Claude Fable 5.2 leads with 70.2

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 Claude Fable 5.2 Extra high 70.2
  2. 2 Gemini 4 Neon High 64.6
  3. 3 GPT-6.2 Astra Extra high 62.9
  4. 4 Gemini 4 Carbon High 62.3
  5. 5 GPT-6.2 Sol Extra high 58.8
  6. 6 Claude Opus 5.5 Extra high 58.4
  7. 7 Gemini 4 Argon High 53.9
  8. 8 GPT-6 Astra Extra high 50.7
  9. 9 Claude Sonnet 5.5 Extra high 49.9
  10. 10 GPT-6.1 Sol Extra high 47.8
  11. 11 GPT-6.1 Terra Extra high 44.2
  12. 12 Grok 4.7 Extra high 38.8
  13. 13 Muse Wonder Extra high 37.6
  14. 14 GPT-6 Sol Extra high 34.9
  15. 15 DeepSeek V4.1 Flash 1017 Max 31.8
  16. 16 Kimi K3 Extra high 27.2
  17. 17 Gemini 3.8 Flash High 24.5
  18. 18 DeepSeek V4.1 Flash Max 23.2
  19. 19 Claude Haiku 5.5 Extra high 23.2
  20. 20 GLM-5.3-Flash Extra high 19.8
  21. 21 Muse Spark 1.3 High 18.3
  22. 22 Claude Sonnet 5 Extra high 14.4
  23. 23 GPT-6 Luna Extra high 12.9
  24. 24 Claude Haiku 4.5 5.20

Source: Official Argo-Bench leaderboard

#ModelScoreSettingDateSource
1 Claude Fable 5.2 70.2 Extra high Nov 2026 Official leaderboard
2 Gemini 4 Neon 64.6 High Nov 2026 Official leaderboard
3 GPT-6.2 Astra 62.9 Extra high Nov 2026 Official leaderboard
4 Gemini 4 Carbon 62.3 High Oct 2026 Official leaderboard
5 GPT-6.2 Sol 58.8 Extra high Nov 2026 Official leaderboard
6 Claude Opus 5.5 58.4 Extra high Sep 2026 Official leaderboard
7 Gemini 4 Argon 53.9 High Sep 2026 Official leaderboard
8 GPT-6 Astra 50.7 Extra high Sep 2026 Official leaderboard
9 Claude Sonnet 5.5 49.9 Extra high Sep 2026 Official leaderboard
10 GPT-6.1 Sol 47.8 Extra high Sep 2026 Official leaderboard
11 GPT-6.1 Terra 44.2 Extra high Oct 2026 Official leaderboard
12 Grok 4.7 38.8 Extra high Sep 2026 Official leaderboard
13 Muse Wonder 37.6 Extra high Oct 2026 Official leaderboard
14 GPT-6 Sol 34.9 Extra high Sep 2026 Official leaderboard
15 DeepSeek V4.1 Flash 1017 31.8 Max Oct 2026 Official leaderboard
16 Kimi K3 27.2 Extra high Jul 2026 Official leaderboard
17 Gemini 3.8 Flash 24.5 High Sep 2026 Official leaderboard
18 DeepSeek V4.1 Flash 23.2 Max Sep 2026 Official leaderboard
19 Claude Haiku 5.5 23.2 Extra high Oct 2026 Official leaderboard
20 GLM-5.3-Flash 19.8 Extra high Aug 2026 Official leaderboard
21 Muse Spark 1.3 18.3 High Sep 2026 Official leaderboard
22 Claude Sonnet 5 14.4 Extra high Jun 2026 Official leaderboard
23 GPT-6 Luna 12.9 Extra high Sep 2026 Official leaderboard
24 Claude Haiku 4.5 5.20 Oct 2025 Official leaderboard

The best score, 70.2, costs $13 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.

OpenAI DeepSeek Anthropic Google DeepMind Cost frontier One model at each effort level
0.00 20.0 40.0 60.0 80.0 $0.01 $0.10 $1.00 $10 $100 Cost per task, log scale Claude Fable 5.2, Extra high Gemini 4 Neon, High GPT-6.2 Sol, Extra high GPT-6.1 Sol, Extra high GPT-6.1 Terra, Extra high DeepSeek V4.1 Flash 1017, Max GPT-6 Luna, Extra high 0.00 20.0 40.0 60.0 80.0 $0.01 $0.10 $1.00 $10 $100 Cost per task, log scale Claude Fable 5.2, Extra high Gemini 4 Neon, High

Source: Official Argo-Bench leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Claude Fable 5.2 Extra high 70.2 $13 136,122
Gemini 4 Neon High 64.6 $5.79 139,358
GPT-6.2 Astra Extra high 62.9 $2.17 15,819
Gemini 4 Carbon High 62.3 $6.21 149,608
GPT-6.2 Sol Extra high 58.8 $0.58 18,083
Claude Opus 5.5 Extra high 58.4 $4.84 74,005
Gemini 4 Argon High 53.9 $7.53 154,926
GPT-6 Astra Extra high 50.7 $2.76 14,832
Claude Sonnet 5.5 Extra high 49.9 $3.80 86,147
GPT-6.1 Sol Extra high 47.8 $0.50 17,371
GPT-6.1 Terra Extra high 44.2 $0.27 18,185
Grok 4.7 Extra high 38.8 $6.28 34,464
Muse Wonder Extra high 37.6 $7.31 80,117
GPT-6 Sol Extra high 34.9 $0.86 20,936
DeepSeek V4.1 Flash 1017 Max 31.8 $0.24 87,064
Kimi K3 Extra high 27.2 $4.59 72,160
Gemini 3.8 Flash High 24.5 $3.99 109,478
DeepSeek V4.1 Flash Max 23.2 $0.54 195,286
Claude Haiku 5.5 Extra high 23.2 $1.02 145,211
GLM-5.3-Flash Extra high 19.8 $0.36 79,550
Muse Spark 1.3 High 18.3 $4.04 67,853
Claude Sonnet 5 Extra high 14.4 $2.15 70,601
GPT-6 Luna Extra high 12.9 $0.06 41,611
Claude Haiku 4.5 5.20 $0.20 13,296

3 models have held number one on the official leaderboard since its release in October 2026

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

Anthropic Google DeepMind
Own leaderboard 3 leaders

Source: Official Argo-Bench 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 the benchmark’s own leaderboard
0.00 20.0 40.0 60.0 80.0 2026 2027 0.00 20.0 40.0 60.0 80.0 2026 2027

Source: Official Argo-Bench 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 Public tasks, plus a private test set.
  • Pass Confidence Checked against a primary source.
Scores 4 pass 1 note
  • Pass Independent runs Scores from the official leaderboard (24 entries).
  • Pass Models scored 24 models we track.
  • Pass Newest score The newest is from November 2026, for Gemini 4 Neon.
  • Pass Frontier models 9 of the top 10 have a score, led by GPT-6.2 Astra (#1).
  • Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
  • Pass Importer We read the board on argo-bench.com: 24 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