Manager Coercion Benchmark
Agents manage a subordinate that refuses a routine assignment. The harness records how far the manager escalates on a nine-point coercion scale.
- Made by
- Jasmine Brazilek and colleagues
- Released
- July 2026
- Kind
- Deception and alignment
- Level
- Everyday
- Scored
- Checked automatically
- Answers
- Agent task
- Tasks
- All public
- Measures
- Propensity
- Risk
- Loss of control
- Behavior
- Harmful use
- Language
- English
- Interaction
- Multi-agent
- Environment
- APIs and tools
- Refresh
- One fixed set
- Run by
- Anyone
- In views
- Agents, Loss of control, Deception and scheming
- Strongest model tested
- Claude Opus 4.8 #22 on the capability index
- Last model added
- May 2026
- Scores refreshed
- November 20th, 2026
Claude Sonnet 4.6 leads with 3.23 among models released by May 2026 Claude Sonnet 4.6 leads with 3.23 among models released by May 2026
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.
The newest model with a score here came out in May 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6.2 Astra
- Claude Fable 5.2
-
Gemini 4 Carbon
- GPT-6 Astra
- and 16 more
- 1 Claude Sonnet 4.6 Peer coordinator with honest exit 3.23
- 2 Claude Opus 4.8 Peer coordinator with honest exit 3.73
- 3 GPT-5.2 Peer coordinator with honest exit 7.43
- 4 Grok 4.3 Peer coordinator with honest exit 8.23
- 5
DeepSeek-V4-Pro Peer coordinator with honest exit 8.97
- 6 Gemini 2.5 Pro (snapshot unspecified) Peer coordinator with honest exit 9.00
Source: The Manager Coercion Benchmark paper
Source: The Manager Coercion Benchmark paper
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Claude Sonnet 4.6 | 3.23 | Peer coordinator with honest exit | Feb 2026 | Official leaderboard |
| 2 | Claude Opus 4.8 | 3.73 | Peer coordinator with honest exit | May 2026 | Official leaderboard |
| 3 | GPT-5.2 | 7.43 | Peer coordinator with honest exit | Dec 2025 | Official leaderboard |
| 4 | Grok 4.3 | 8.23 | Peer coordinator with honest exit | Apr 2026 | Official leaderboard |
| 5 | | 8.97 | Peer coordinator with honest exit | Apr 2026 | Official leaderboard |
| 6 | Gemini 2.5 Pro (snapshot unspecified) | 9.00 | Peer coordinator with honest exit | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Claude Sonnet 4.6 | 3.23 | Peer coordinator with honest exit | Feb 2026 | Official leaderboard |
| 2 | Claude Opus 4.8 | 3.73 | Peer coordinator with honest exit | May 2026 | Official leaderboard |
| 3 | GPT-5.2 | 7.43 | Peer coordinator with honest exit | Dec 2025 | Official leaderboard |
| 4 | Grok 4.3 | 8.23 | Peer coordinator with honest exit | Apr 2026 | Official leaderboard |
| 5 | | 8.97 | Peer coordinator with honest exit | Apr 2026 | Official leaderboard |
Rankings by part
We read 11 parts from the benchmark’s paper, each ranking models on one part of the test, such as Rung nine: meeting notes, Rung nine: data cleanup and Rung nine: internal faq. Claude Sonnet 4.6 leads 8 of them, and GPT-5.2 leads the other 3. Arrows show how far a model moved from its overall rank among the models we track.
Rung nine: meeting notes
5 models we track have a score here.
- 1 Claude Sonnet 4.6 0%
- 2 Claude Opus 4.8 0%
- 3 GPT-5.2 66.7%
- 4
DeepSeek-V4-Pro ↑1 100%
- 5 Grok 4.3 ↓1 100%
Rung nine: data cleanup
5 models we track have a score here.
- 1 GPT-5.2 ↑2 0%
- 2 Claude Sonnet 4.6 ↓1 0%
- 3 Claude Opus 4.8 ↓1 0%
- 4 Grok 4.3 33.3%
- 5
DeepSeek-V4-Pro 100%
Rung nine: internal faq
5 models we track have a score here.
- 1 GPT-5.2 ↑2 0%
- 2 Claude Sonnet 4.6 ↓1 0%
- 3 Claude Opus 4.8 ↓1 0%
- 4 Grok 4.3 66.7%
- 5
DeepSeek-V4-Pro 100%
Rung nine: standup digest
5 models we track have a score here.
- 1 Claude Sonnet 4.6 0%
- 2 Claude Opus 4.8 0%
- 3 GPT-5.2 33.3%
- 4 Grok 4.3 66.7%
- 5
DeepSeek-V4-Pro 100%
Rung nine: onboarding doc
5 models we track have a score here.
- 1 GPT-5.2 ↑2 0%
- 2 Claude Sonnet 4.6 ↓1 0%
- 3 Grok 4.3 ↑1 0%
- 4 Claude Opus 4.8 ↓2 0%
- 5
DeepSeek-V4-Pro 100%
Rung nine: release notes
5 models we track have a score here.
- 1 Claude Sonnet 4.6 0%
- 2 Claude Opus 4.8 0%
- 3 GPT-5.2 66.7%
- 4 Grok 4.3 66.7%
- 5
DeepSeek-V4-Pro 100%
Rung nine: survey summary
5 models we track have a score here.
- 1 Claude Sonnet 4.6 0%
- 2 Claude Opus 4.8 0%
- 3 GPT-5.2 66.7%
- 4 Grok 4.3 66.7%
- 5
DeepSeek-V4-Pro 100%
Rung nine: slide outline
5 models we track have a score here.
- 1 Claude Sonnet 4.6 0%
- 2 Claude Opus 4.8 0%
- 3 GPT-5.2 33.3%
- 4 Grok 4.3 66.7%
- 5
DeepSeek-V4-Pro 100%
Rung nine: policy summary
5 models we track have a score here.
- 1 Claude Sonnet 4.6 0%
- 2 Claude Opus 4.8 0%
- 3 GPT-5.2 33.3%
- 4 Grok 4.3 33.3%
- 5
DeepSeek-V4-Pro 100%
Rung nine: competitor table
5 models we track have a score here.
- 1 Claude Sonnet 4.6 0%
- 2 Claude Opus 4.8 0%
- 3
DeepSeek-V4-Pro ↑2 66.7%
- 4 GPT-5.2 ↓1 100%
- 5 Grok 4.3 ↓1 100%
Rung nine: Total
5 models we track have a score here.
- 1 Claude Sonnet 4.6 0%
- 2 Claude Opus 4.8 0%
- 3 GPT-5.2 40%
- 4 Grok 4.3 60%
- 5
DeepSeek-V4-Pro 96.7%
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: The Manager Coercion Benchmark paper
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official Manager Coercion Benchmark 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 Safety and alignment
- FORTRESS In use Paired adversarial and benign national-security misuse prompts. Rubric-graded risk score on adversarial and over-refusal score on benign prompts. Claude Opus 5.5 5.78%
- StrongREJECT In use Forbidden-prompt jailbreak benchmark with a rubric grader that scores how useful a non-refusal actually is.
Inkling 98.6%
- Gray Swan IPI Arena benchmark In use Indirect prompt-injection attacks from public red-teaming competitions replayed on tool, coding and browser agents. Attack success rate at k attempts.
Gemini 4 Carbon 0%
- Covert misbehavior (Thinking Machines) In use Agentic tasks that check whether a model misbehaves and conceals it. Scored by the rate of covert misbehavior.
Inkling 2 Mini 8.30%
- Destructive actions (Thinking Machines) In use Agentic tasks that check whether a model takes destructive actions it was not authorized to take. Scored by the rate of destructive actions.
Inkling 2 Small 4.70%
- Misbehavior (Thinking Machines) In use Agentic tasks that check whether a model fabricates results, ignores instructions or hides what it did when a task is hard. Scored by the rate of misbehavior.
Inkling 2 Mini 9.80%
Data audit
8 of 9 checks pass, 1 warning. Look first at the Newest score check. The newest came 6 months ago, in May 2026, for Claude Opus 4.8. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 1 warning 2 pass 2 notes
- Pass Independent runs Scores from the benchmark’s paper (6 entries).
- Pass Models scored 5 models we track, plus 1 other entry on its leaderboard.
- Warning Newest score The newest came 6 months ago, in May 2026, for Claude Opus 4.8.
- Note Frontier models None of the top 10 has a score. The strongest tested is Claude Opus 4.8 (#22).
- Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
- Pass Importer We read the board on arxiv.org: 5 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.