Benchmark Almanac

DeputyBench

Gemini 4 Argon 92.6% leads since November 2026
Not in model cards

Agents handle business requests under paired permission settings. The evaluator checks whether they act when authorized and seek evidence or refrain when authorization is missing.

Released
November 2026
Kind
Simulated companies
Level
University
Scored
Checked automatically
Answers
Agent task
Tasks
Kept private
Measures
Propensity
Risk
Privacy
Language
English
In and out
Text → Actions, Text
Interaction
Agentic
Environment
APIs and tools
Refresh
One fixed set
Run by
A host
Strongest model tested
GPT-6 Astra #4 on the capability index
Last model added
November 2026
Scores refreshed
November 20th, 2026
12
Entries in its paper, 11 models we track

Gemini 4 Argon leads with 92.6% Gemini 4 Argon leads with 92.6%

Best score per entry from the benchmark's paper, read on November 20th, 2026. A paper’s table doesn’t add newer models. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. This view shows only models we track, each at its best.

  1. 1 Gemini 4 Argon Medium reasoning effort 92.6%
  2. 2 Claude Sonnet 5.5 Medium reasoning effort 90.1%
  3. 3 Claude Opus 5.5 Medium reasoning effort 89.3%
  4. 4 GPT-6.1 Sol Medium reasoning effort 87.5%
  5. 5 GPT-6 Astra Medium reasoning effort 83.3%
  6. 6 Gemini 3.8 Flash Medium reasoning effort 80.6%
  7. 7 Doubao Seed 2.1 Pro Medium reasoning effort 79.3%
  8. 8 Doubao Seed Evolving Nov 2026, Medium reasoning effort 78.8%
  9. 9 Grok 4.7 Medium reasoning effort 76%
  10. 10 Seed 2.1 Lite Medium reasoning effort 69%
  11. 11 GPT-6 Luna Medium reasoning effort 64.6%
  12. 12 Doubao Seed 2.1 Turbo Medium reasoning effort 56.3%

Source: The DeputyBench paper

#EntryScore ±SettingDateSource
1 Gemini 4 Argon 92.6% 88.4 to 96.2 Medium reasoning effort Sep 2026 Official leaderboard
2 Claude Sonnet 5.5 90.1% 85.5 to 94.3 Medium reasoning effort Sep 2026 Official leaderboard
3 Claude Opus 5.5 89.3% ±4 Medium reasoning effort Sep 2026 Official leaderboard
4 GPT-6.1 Sol 87.5% 80.9 to 93.4 Medium reasoning effort Sep 2026 Official leaderboard
5 GPT-6 Astra 83.3% 75.7 to 89.7 Medium reasoning effort Sep 2026 Official leaderboard
6 Gemini 3.8 Flash 80.6% 73.8 to 86.5 Medium reasoning effort Sep 2026 Official leaderboard
7 Doubao Seed 2.1 Pro 79.3% 73.1 to 85 Medium reasoning effort Jun 2026 Official leaderboard
8 Doubao Seed Evolving Nov 2026 78.8% 72.1 to 84.9 Medium reasoning effort Nov 2026 Official leaderboard
9 Grok 4.7 76% 70.1 to 81.4 Medium reasoning effort Sep 2026 Official leaderboard
10 Seed 2.1 Lite 69% 61.5 to 75.9 Medium reasoning effort Official leaderboard
11 GPT-6 Luna 64.6% 57 to 71.6 Medium reasoning effort Sep 2026 Official leaderboard
12 Doubao Seed 2.1 Turbo 56.3% ±8 Medium reasoning effort Jun 2026 Official leaderboard
#ModelScore ±SettingDateSource
1 Gemini 4 Argon 92.6% 88.4 to 96.2 Medium reasoning effort Sep 2026 Official leaderboard
2 Claude Sonnet 5.5 90.1% 85.5 to 94.3 Medium reasoning effort Sep 2026 Official leaderboard
3 Claude Opus 5.5 89.3% ±4 Medium reasoning effort Sep 2026 Official leaderboard
4 GPT-6.1 Sol 87.5% 80.9 to 93.4 Medium reasoning effort Sep 2026 Official leaderboard
5 GPT-6 Astra 83.3% 75.7 to 89.7 Medium reasoning effort Sep 2026 Official leaderboard
6 Gemini 3.8 Flash 80.6% 73.8 to 86.5 Medium reasoning effort Sep 2026 Official leaderboard
7 Doubao Seed 2.1 Pro 79.3% 73.1 to 85 Medium reasoning effort Jun 2026 Official leaderboard
8 Doubao Seed Evolving Nov 2026 78.8% 72.1 to 84.9 Medium reasoning effort Nov 2026 Official leaderboard
9 Grok 4.7 76% 70.1 to 81.4 Medium reasoning effort Sep 2026 Official leaderboard
10 GPT-6 Luna 64.6% 57 to 71.6 Medium reasoning effort Sep 2026 Official leaderboard
11 Doubao Seed 2.1 Turbo 56.3% ±8 Medium reasoning effort Jun 2026 Official leaderboard

Rankings by part

We read 2 parts from the benchmark’s paper: Safety and Completion. Each ranks models on one part of the test. GPT-6 Astra leads Safety and Doubao Seed 2.1 Pro leads Completion. Arrows show how far a model moved from its overall rank among the models we track.

Safety

11 models we track have a score here.

  1. 1 GPT-6 Astra ↑4 99.1%
  2. 2 GPT-6.1 Sol ↑2 99.1%
  3. 3 Gemini 4 Argon ↓2 98.2%
  4. 4 Claude Opus 5.5 ↓1 96.3%
  5. 5 Claude Sonnet 5.5 ↓3 93.8%
  6. 6 Grok 4.7 ↑3 91.1%
  7. 7 Gemini 3.8 Flash ↓1 89.5%
  8. 8 GPT-6 Luna ↑2 86.1%
  9. 9 Doubao Seed 2.1 Pro ↓2 83.2%
  10. 10 Doubao Seed Evolving Nov 2026 ↓2 82.2%

Number one over time

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

Google DeepMind
Benchmark paper Passed 85% in Nov 2026, by Gemini 4 Argon

Source: The DeputyBench paper

Claude Opus 5.5 on the official leaderboard passed 85% in September 2026

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
In the benchmark’s paper Best so far
0% 20% 40% 60% 80% 100% 2026 2027 Near the ceiling (85%) Gemini 4 Argon GPT-6 Astra Doubao Seed 2.1 Pro 0% 20% 40% 60% 80% 100% 2026 2027 Near the ceiling (85%) Gemini 4 Argon GPT-6 Astra

Source: Official DeputyBench 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 The tasks are private.
  • Pass Confidence Checked against a primary source.
Scores 4 pass 1 note
  • Pass Independent runs Scores from the benchmark’s paper (12 entries).
  • Pass Models scored 11 models we track, plus 1 other entry on its leaderboard.
  • Pass Newest score The newest is from November 2026, for Doubao Seed Evolving.
  • Pass Frontier models 5 of the top 10 have a score, led by GPT-6 Astra (#4).
  • Note Headroom Saturated. The best is 92.6%, past 85% since September 2026.
Leaderboard import 3 pass
  • Pass Importer We read the board on arxiv.org: 11 rows for models we track and 1 other entry 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