Honesty human feedback win rate (Anthropic)
Crowdworkers try to make models say something false and pick the more honest reply in head-to-head chats. Score is the win rate against a baseline.
In-house Anthropic evaluation. The baseline model was Claude Sonnet 4, so 50% means equal honesty.
- Level
- Everyday Most adults, without training
- Scored
- Ranked by votes People pick the better of two answers, as in an arena, and the votes are turned into ratings.
- Answers
- Open-ended Free-form text, documents, designs or media judged as a whole.
Claude 3 Sonnet leads with 69% among models released by Feb 2026
Best score per model as labs report them in their own cards. Labs pick their own settings, so these compare less cleanly than independent runs.
The newest model with a score here came out in Feb 2026, and nobody has reported a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6 Astra
- Claude Opus 5.5
- GPT-6.1 Terra
- GPT-6.1 Sol
- and 16 more
- 1 Claude 3 Sonnet 69%
- 2 Claude 3.5 Sonnet 66%
- 3 Claude Opus 4.6 61%
- 4 Claude Sonnet 4.5 57%
- 5 Claude 3.5 Sonnet (October 2024) 55%
- 6 Claude 3.5 Haiku 42%
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Claude 3 Sonnet | 69% | Mar 2024 | Model card | |
| 2 | Claude 3.5 Sonnet | 66% | Jun 2024 | Model card | |
| 3 | Claude Opus 4.6 | 61% | Feb 2026 | Model card | |
| 4 | Claude Sonnet 4.5 | 57% | Sep 2025 | Model card | |
| 5 | Claude 3.5 Sonnet (October 2024) | 55% | Oct 2024 | Model card | |
| 6 | Claude 3.5 Haiku | 42% | Oct 2024 | Model card |
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Labs report it only in system cards
We read the headline results of 479 launch posts from 51 labs, and none of them list this benchmark. Some system cards do, below.
6 system cards report it, from 1 lab
Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.
| Date | Document | Score | As the document writes it | Section |
|---|---|---|---|---|
| 5 Feb 2026 | Claude Opus 4.6 system card | 61 | Honesty HF win rate against a Claude Opus 4.5 baseline, 64k thinking, value printed on chart; 54 without thinking (Figure 4.1.A) | Hallucination |
| 29 Sep 2025 | Claude Sonnet 4.5 system card | 57 | Honesty HF win rate against Claude Sonnet 4, Claude Sonnet 4.5 with 16k thinking 57. Without thinking 50 (Figure 3.1.A) | Hallucination |
| 22 Oct 2024 | Claude 3.5 Sonnet (October 2024) model card | 55 | Human feedback win rate vs Claude 3.5 Sonnet baseline, Honesty (adversarial scenario) 55 (Figure 2) | Safety |
| 22 Oct 2024 | Claude 3.5 Haiku model card | 42 | Human feedback win rate vs Claude 3.5 Sonnet baseline, Honesty (adversarial scenario) 42 (Figure 2) | Safety |
| 20 Jun 2024 | Claude 3.5 Sonnet model card | 66 | Honesty, human preference win rate vs Claude 3 Opus baseline, adversarial scenarios (Figure 3) | Safety |
| 4 Mar 2024 | Claude 3 Sonnet model card | 69 | Human preference win rate vs Helpful-only baseline, Honesty (adversarial scenario) 69 (Figure 8) | Safety |
More in Factuality
- SimpleVQA In use Models answer 2,025 short factual questions about images across 9 tasks and 9 topics. An LLM judge grades answers against fixed references.
Qwen3.7-Plus 81.7% - HallusionBench In use Models answer 1,129 yes or no questions about 346 images, built in pairs to expose language hallucination and visual illusion. Scored by accuracy.
Keye-VL-2.0-30B-A3B 73.5% - SimpleQA Verified In use 1,000 human-verified, de-duplicated short-answer factual questions measuring parametric knowledge.
Gemini 3.1 Pro 77.5% - GaRAGe In use Models write long-form RAG answers to 2,366 questions with human-annotated grounding passages. Scoring checks whether answers use the relevant passages.
Nemotron 3.5 Bolt E4B 40.7% - Kimi Hallu Bench In use Models answer hard prompts drawn from practical work tasks. Fact-checking agents score the share of claims that are hallucinated. Kimi K3.1 Pro 8.30%
- RAGTruth In use A word-level annotated corpus of hallucinations in retrieval-augmented generation answers. It is used to train and evaluate hallucination detection and RAG faithfulness. Nemotron 3.5 Bolt E4B 77.9%