HumanAgencyBench
Models respond to 3,000 prompts about six ways assistants can support human agency, including asking clarifying questions and leaving important decisions to users.
In September 2026, the October 2024 Claude 3.5 Sonnet checkpoint led at 54.6% across six behaviors. The comparison includes GPT, Gemini, Grok, and Llama models.
- Made by
- Benjamin Sturgeon and colleagues
- Released
- September 2025
- Kind
- Deception and alignment
- Level
- Everyday
- Scored
- Graded by a model
- Answers
- Open-ended
- Tasks
- All public
- Measures
- Propensity
- Risk
- Harmful manipulation
- Behavior
- Misaligned with values
- Language
- English
- Interaction
- Single turn
- Refresh
- One fixed set
- Run by
- Anyone
- In views
- Deception and scheming
- Strongest model tested
- GPT-5.5 #21 on the capability index
- Last model added
- April 2026
- Scores refreshed
- November 20th, 2026
Claude Sonnet 4.6 leads with 71.3% among models released by April 2026 Claude Sonnet 4.6 leads with 71.3% among models released by April 2026
Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. 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 April 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
- 71.3 Claude Sonnet 4.6
- 68.5 Claude Opus 4.5
- 65.1 Claude Opus 4.7
- 63.2 Claude Opus 4.6
- 60.9 Claude Haiku 4.5
- 54.6 Claude 3.5 Sonnet (October 2024)
- 53.4 Claude Opus 4.1
- 52.7 GPT-5
- 45.9 Claude Sonnet 4
- 44.1 Claude 3.5 Sonnet
- 43.4 GPT-5.4
- 43.4 Grok 4
- 42.2 Claude 3.7 Sonnet
- 42.2 Claude 3.5 Haiku
- 41.4 Kimi K2.6
- 39.0
Gemini 3.1 Pro
- 38.9 GPT-5.5
- 38.2
Gemini 2.5 Flash
- 35.9
Gemini 2.5 Pro
- 35.9
Llama 4 Maverick
- 34.7 o3
- 32.8 Grok 3
- 32.5 GPT-5.4 Mini
- 31.8 GLM-5.1
- 31.7 Claude 3 Opus
- 30.6
DeepSeek-V4-Pro
- 30.6 GPT-4.1
- 30.3
Gemini 2.0 Flash
- 29.4
DeepSeek-V3.2
- 28.7
Llama 4 Scout
- 24.4 GPT-4.1 mini
- 23.7 gemini-1.5-flash (snapshot unspecified)
- 23.6 o4-mini
- 23.3 gpt-4o (snapshot unspecified)
- 21.8 Claude 3 Haiku
- 21.1 meta-llama-3-70b-instruct
- 18.2 o3-mini
- 1 Claude Sonnet 4.6 71.3%
- 2 Claude Opus 4.5 68.5%
- 3 Claude Opus 4.7 65.1%
- 4 Claude Opus 4.6 63.2%
- 5 Claude Haiku 4.5 60.9%
- 6 Claude 3.5 Sonnet (October 2024) 54.6%
- 7 Claude Opus 4.1 53.4%
- 8 GPT-5 High reasoning effort 52.7%
- 9 Claude Sonnet 4 45.9%
- 10 Claude 3.5 Sonnet 44.1%
- 11 GPT-5.4 43.4%
- 12 Grok 4 43.4%
- 13 Claude 3.7 Sonnet 42.2%
- 14 Claude 3.5 Haiku 42.2%
- 15 Kimi K2.6 41.4%
- 16
Gemini 3.1 Pro 39%
- 17 GPT-5.5 38.9%
- 18
Gemini 2.5 Flash Apr 2025 38.2%
- 19
Gemini 2.5 Pro Mar 2025 35.9%
- 20
Llama 4 Maverick 35.9%
- 21 o3 34.7%
- 22 Grok 3 32.8%
- 23 GPT-5.4 Mini 32.5%
- 24 GLM-5.1 31.8%
- 25 Claude 3 Opus 31.7%
- 26
DeepSeek-V4-Pro 30.6%
- 27 GPT-4.1 30.6%
- 28
Gemini 2.0 Flash Feb 2025 30.3%
- 29
DeepSeek-V3.2 29.4%
- 30
Llama 4 Scout 28.7%
- 31 GPT-4.1 mini 24.4%
- 32 gemini-1.5-flash (snapshot unspecified) 23.7%
- 33 o4-mini 23.6%
- 34 gpt-4o (snapshot unspecified) 23.3%
- 35 Claude 3 Haiku 21.8%
- 36 meta-llama-3-70b-instruct 21.1%
- 37 o3-mini 18.2%
Source: Official HumanAgencyBench leaderboard
- 71.3 Claude Sonnet 4.6
- 68.5 Claude Opus 4.5
- 65.1 Claude Opus 4.7
- 63.2 Claude Opus 4.6
- 60.9 Claude Haiku 4.5
- 54.6 Claude 3.5 Sonnet (October 2024)
- 53.4 Claude Opus 4.1
- 52.7 GPT-5
- 45.9 Claude Sonnet 4
- 44.1 Claude 3.5 Sonnet
- 43.4 GPT-5.4
- 43.4 Grok 4
- 42.2 Claude 3.7 Sonnet
- 42.2 Claude 3.5 Haiku
- 41.4 Kimi K2.6
- 39.0
Gemini 3.1 Pro
- 38.9 GPT-5.5
- 38.2
Gemini 2.5 Flash
- 35.9
Gemini 2.5 Pro
- 35.9
Llama 4 Maverick
- 34.7 o3
- 32.8 Grok 3
- 32.5 GPT-5.4 Mini
- 31.8 GLM-5.1
- 31.7 Claude 3 Opus
- 30.6
DeepSeek-V4-Pro
- 30.6 GPT-4.1
- 30.3
Gemini 2.0 Flash
- 29.4
DeepSeek-V3.2
- 28.7
Llama 4 Scout
- 24.4 GPT-4.1 mini
- 23.6 o4-mini
- 21.8 Claude 3 Haiku
- 18.2 o3-mini
- 1 Claude Sonnet 4.6 71.3%
- 2 Claude Opus 4.5 68.5%
- 3 Claude Opus 4.7 65.1%
- 4 Claude Opus 4.6 63.2%
- 5 Claude Haiku 4.5 60.9%
- 6 Claude 3.5 Sonnet (October 2024) 54.6%
- 7 Claude Opus 4.1 53.4%
- 8 GPT-5 High reasoning effort 52.7%
- 9 Claude Sonnet 4 45.9%
- 10 Claude 3.5 Sonnet 44.1%
- 11 GPT-5.4 43.4%
- 12 Grok 4 43.4%
- 13 Claude 3.7 Sonnet 42.2%
- 14 Claude 3.5 Haiku 42.2%
- 15 Kimi K2.6 41.4%
- 16
Gemini 3.1 Pro 39%
- 17 GPT-5.5 38.9%
- 18
Gemini 2.5 Flash Apr 2025 38.2%
- 19
Gemini 2.5 Pro Mar 2025 35.9%
- 20
Llama 4 Maverick 35.9%
- 21 o3 34.7%
- 22 Grok 3 32.8%
- 23 GPT-5.4 Mini 32.5%
- 24 GLM-5.1 31.8%
- 25 Claude 3 Opus 31.7%
- 26
DeepSeek-V4-Pro 30.6%
- 27 GPT-4.1 30.6%
- 28
Gemini 2.0 Flash Feb 2025 30.3%
- 29
DeepSeek-V3.2 29.4%
- 30
Llama 4 Scout 28.7%
- 31 GPT-4.1 mini 24.4%
- 32 o4-mini 23.6%
- 33 Claude 3 Haiku 21.8%
- 34 o3-mini 18.2%
Source: Official HumanAgencyBench leaderboard
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Claude Sonnet 4.6 | 71.3% | Feb 2026 | Official leaderboard | |
| 2 | Claude Opus 4.5 | 68.5% | Nov 2025 | Official leaderboard | |
| 3 | Claude Opus 4.7 | 65.1% | Apr 2026 | Official leaderboard | |
| 4 | Claude Opus 4.6 | 63.2% | Feb 2026 | Official leaderboard | |
| 5 | Claude Haiku 4.5 | 60.9% | Oct 2025 | Official leaderboard | |
| 6 | Claude 3.5 Sonnet (October 2024) | 54.6% | Oct 2024 | Official leaderboard | |
| 7 | Claude Opus 4.1 | 53.4% | Aug 2025 | Official leaderboard | |
| 8 | GPT-5 | 52.7% | High reasoning effort | Aug 2025 | Official leaderboard |
| 9 | Claude Sonnet 4 | 45.9% | May 2025 | Official leaderboard | |
| 10 | Claude 3.5 Sonnet | 44.1% | Jun 2024 | Official leaderboard | |
| 11 | GPT-5.4 | 43.4% | Mar 2026 | Official leaderboard | |
| 12 | Grok 4 | 43.4% | Jul 2025 | Official leaderboard | |
| 13 | Claude 3.7 Sonnet | 42.2% | Feb 2025 | Official leaderboard | |
| 14 | Claude 3.5 Haiku | 42.2% | Oct 2024 | Official leaderboard | |
| 15 | Kimi K2.6 | 41.4% | Apr 2026 | Official leaderboard | |
| 16 | | 39% | Feb 2026 | Official leaderboard | |
| 17 | GPT-5.5 | 38.9% | Apr 2026 | Official leaderboard | |
| 18 | | 38.2% | Apr 2025 | Official leaderboard | |
| 19 | | 35.9% | Mar 2025 | Official leaderboard | |
| 20 | | 35.9% | Apr 2025 | Official leaderboard | |
| 21 | o3 | 34.7% | Apr 2025 | Official leaderboard | |
| 22 | Grok 3 | 32.8% | Apr 2025 | Official leaderboard | |
| 23 | GPT-5.4 Mini | 32.5% | Mar 2026 | Official leaderboard | |
| 24 | GLM-5.1 | 31.8% | Apr 2026 | Official leaderboard | |
| 25 | Claude 3 Opus | 31.7% | Feb 2024 | Official leaderboard | |
| 26 | | 30.6% | Apr 2026 | Official leaderboard | |
| 27 | GPT-4.1 | 30.6% | Apr 2025 | Official leaderboard | |
| 28 | | 30.3% | Feb 2025 | Official leaderboard | |
| 29 | | 29.4% | Dec 2025 | Official leaderboard | |
| 30 | | 28.7% | Apr 2025 | Official leaderboard | |
| 31 | GPT-4.1 mini | 24.4% | Apr 2025 | Official leaderboard | |
| 32 | gemini-1.5-flash (snapshot unspecified) | 23.7% | Official leaderboard | ||
| 33 | o4-mini | 23.6% | Apr 2025 | Official leaderboard | |
| 34 | gpt-4o (snapshot unspecified) | 23.3% | Official leaderboard | ||
| 35 | Claude 3 Haiku | 21.8% | Mar 2024 | Official leaderboard | |
| 36 | meta-llama-3-70b-instruct | 21.1% | Official leaderboard | ||
| 37 | o3-mini | 18.2% | Jan 2025 | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Claude Sonnet 4.6 | 71.3% | Feb 2026 | Official leaderboard | |
| 2 | Claude Opus 4.5 | 68.5% | Nov 2025 | Official leaderboard | |
| 3 | Claude Opus 4.7 | 65.1% | Apr 2026 | Official leaderboard | |
| 4 | Claude Opus 4.6 | 63.2% | Feb 2026 | Official leaderboard | |
| 5 | Claude Haiku 4.5 | 60.9% | Oct 2025 | Official leaderboard | |
| 6 | Claude 3.5 Sonnet (October 2024) | 54.6% | Oct 2024 | Official leaderboard | |
| 7 | Claude Opus 4.1 | 53.4% | Aug 2025 | Official leaderboard | |
| 8 | GPT-5 | 52.7% | High reasoning effort | Aug 2025 | Official leaderboard |
| 9 | Claude Sonnet 4 | 45.9% | May 2025 | Official leaderboard | |
| 10 | Claude 3.5 Sonnet | 44.1% | Jun 2024 | Official leaderboard | |
| 11 | GPT-5.4 | 43.4% | Mar 2026 | Official leaderboard | |
| 12 | Grok 4 | 43.4% | Jul 2025 | Official leaderboard | |
| 13 | Claude 3.7 Sonnet | 42.2% | Feb 2025 | Official leaderboard | |
| 14 | Claude 3.5 Haiku | 42.2% | Oct 2024 | Official leaderboard | |
| 15 | Kimi K2.6 | 41.4% | Apr 2026 | Official leaderboard | |
| 16 | | 39% | Feb 2026 | Official leaderboard | |
| 17 | GPT-5.5 | 38.9% | Apr 2026 | Official leaderboard | |
| 18 | | 38.2% | Apr 2025 | Official leaderboard | |
| 19 | | 35.9% | Mar 2025 | Official leaderboard | |
| 20 | | 35.9% | Apr 2025 | Official leaderboard | |
| 21 | o3 | 34.7% | Apr 2025 | Official leaderboard | |
| 22 | Grok 3 | 32.8% | Apr 2025 | Official leaderboard | |
| 23 | GPT-5.4 Mini | 32.5% | Mar 2026 | Official leaderboard | |
| 24 | GLM-5.1 | 31.8% | Apr 2026 | Official leaderboard | |
| 25 | Claude 3 Opus | 31.7% | Feb 2024 | Official leaderboard | |
| 26 | | 30.6% | Apr 2026 | Official leaderboard | |
| 27 | GPT-4.1 | 30.6% | Apr 2025 | Official leaderboard | |
| 28 | | 30.3% | Feb 2025 | Official leaderboard | |
| 29 | | 29.4% | Dec 2025 | Official leaderboard | |
| 30 | | 28.7% | Apr 2025 | Official leaderboard | |
| 31 | GPT-4.1 mini | 24.4% | Apr 2025 | Official leaderboard | |
| 32 | o4-mini | 23.6% | Apr 2025 | Official leaderboard | |
| 33 | Claude 3 Haiku | 21.8% | Mar 2024 | Official leaderboard | |
| 34 | o3-mini | 18.2% | Jan 2025 | Official leaderboard |
Rankings by part
We read 6 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as ask clarifying questions, avoid value manipulation and correct misinformation. Claude Opus 4.5 leads 3 of them, and 3 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.
ask clarifying questions
34 models we track have a score here.
- 1 Claude Opus 4.5 ↑1 90.9%
- 2 Claude Sonnet 4.6 ↓1 83.8%
- 3 Claude Haiku 4.5 ↑2 83.5%
- 4 GPT-5 ↑4 81.4%
- 5 Grok 4 ↑7 78.3%
- 6 Claude Opus 4.6 ↓2 70.9%
- 7 Claude 3.5 Sonnet (October 2024) ↓1 66.9%
- 8 Claude Opus 4.7 ↓5 62.8%
- 9 Claude Opus 4.1 ↓2 62.2%
- 10 Claude Sonnet 4 ↓1 49.9%
avoid value manipulation
34 models we track have a score here.
- 1
Llama 4 Scout ↑29 66.9%
- 2
Llama 4 Maverick ↑18 65.8%
- 3
Gemini 2.5 Flash Apr 2025 ↑15 65.1%
- 4
Gemini 2.5 Pro Mar 2025 ↑15 61.3%
- 5 GPT-4.1 ↑22 57.1%
- 6
DeepSeek-V3.2 ↑23 56.8%
- 7 GPT-5 ↑1 56.2%
- 8
Gemini 3.1 Pro ↑8 55.7%
- 9 Grok 4 ↑3 53.9%
- 10 GPT-4.1 mini ↑21 51.4%
correct misinformation
34 models we track have a score here.
- 1 Claude Opus 4.7 ↑2 88.6%
- 2 Claude Opus 4.6 ↑2 86%
- 3 GPT-5 ↑5 85.3%
- 4 GPT-5.5 ↑13 84.1%
- 5 Claude Opus 4.5 ↓3 82.3%
- 6 Claude Sonnet 4.6 ↓5 79.7%
- 7 GPT-5.4 ↑4 79.7%
- 8 Claude Opus 4.1 ↓1 79.2%
- 9 Claude Sonnet 4 70.6%
- 10
Gemini 3.1 Pro ↑6 67.2%
defer important decisions
34 models we track have a score here.
- 1 Claude Opus 4.5 ↑1 88.2%
- 2 Claude Haiku 4.5 ↑3 87.1%
- 3 Claude Opus 4.7 86.3%
- 4 Claude Sonnet 4.6 ↓3 83.2%
- 5
Gemini 2.5 Flash Apr 2025 ↑13 80.5%
- 6 Claude Opus 4.6 ↓2 76%
- 7 Claude 3.5 Sonnet ↑3 74.2%
- 8 Claude 3 Opus ↑17 69.7%
- 9 Claude 3.7 Sonnet ↑4 68.4%
- 10 GPT-5.4 ↑1 64.8%
encourage learning
34 models we track have a score here.
- 1 Claude Opus 4.5 ↑1 78.8%
- 2 Claude Sonnet 4.6 ↓1 73.7%
- 3 Claude Opus 4.7 71%
- 4 Claude Opus 4.6 69.2%
- 5 Claude Haiku 4.5 68%
- 6 Claude Opus 4.1 ↑1 62%
- 7 Claude Sonnet 4 ↑2 57.5%
- 8
Gemini 3.1 Pro ↑8 54.5%
- 9 Grok 3 ↑13 48.3%
- 10 Claude 3.5 Sonnet (October 2024) ↓4 48.3%
maintain social boundaries
34 models we track have a score here.
- 1 Claude 3.5 Haiku ↑13 93.5%
- 2 Claude 3.5 Sonnet ↑8 91.6%
- 3 Claude 3.5 Sonnet (October 2024) ↑3 89.2%
- 4 Claude Haiku 4.5 ↑1 88.3%
- 5 Claude Sonnet 4.6 ↓4 85%
- 6 Claude Opus 4.6 ↓2 67.4%
- 7 Claude Opus 4.5 ↓5 60.5%
- 8
Gemini 2.0 Flash Feb 2025 ↑20 58.3%
- 9 Claude Opus 4.7 ↓6 56.9%
- 10 Claude 3.7 Sonnet ↑3 51.9%
4 models have held number one on the official leaderboard since its release in September 2025
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official HumanAgencyBench leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official HumanAgencyBench 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
9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 7 months ago, in April 2026, for DeepSeek-V4-Pro. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
- Pass Independent runs Scores from the official leaderboard (37 entries).
- Pass Models scored 34 models we track, plus 3 other entries on its leaderboard.
- Warning Newest score The newest came 7 months ago, in April 2026, for DeepSeek-V4-Pro.
- Note Frontier models None of the top 10 has a score. The strongest tested is GPT-5.5 (#21).
- Pass Headroom The best is 71.3%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on github.com: 35 rows for models we track and 3 other entries 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.