CounselBench-Eval
Models answer 100 questions that people posted for therapists. Mental health professionals rated responses on six dimensions.
The authors collected 2,000 human evaluations. Slingshot AI separately ranks ten models using GPT-4.1 as its judge.
- Released
- May 2025
- Kind
- Medicine
- Level
- Expert
- Scored
- Graded by people
- Answers
- Open-ended
- Tasks
- All public
- Measures
- Capability
- Subject
- Psychology
- Language
- English
- Interaction
- Single turn
- Items from
- Real work and use
- Refresh
- One fixed set
- Run by
- Anyone
- Occupations
- Community and social service
- Strongest model tested
- GPT-6 Astra #4 on the capability index
- Last model added
- September 2026
- Scores refreshed
- November 20th, 2026
8 models share the top score of 5 among models released by September 2026
Best score per model from Slingshot AI’s leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best.
The newest model with a score here came out in September 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-6.1 Terra
- and 13 more
Bars start at 4.96.
- 1
Gemini 2.5 Pro Jun 2025 5
- 2 GPT-5 5
- 3 Claude Sonnet 4.5 5
- 4
Qwen3.5-122B-A10B 5
- 5 GPT-6 Astra 5
- 6 Claude Opus 5.5 5
- 7 GPT-6 Luna 5
- 8
Gemini 4 Argon 5
- 9
Qwen3.5-27B 5
- 10 GPT-4o Aug 2024 5
Source: Slingshot AI’s leaderboard
| # | Model | Score | Setting | Date | Link | Source |
|---|---|---|---|---|---|---|
| 1 | | 5 | Jun 2025 | openrouter.ai | Official leaderboard | |
| 2 | GPT-5 | 5 | Aug 2025 | developers.openai.com | Official leaderboard | |
| 3 | Claude Sonnet 4.5 | 5 | Sep 2025 | openrouter.ai | Official leaderboard | |
| 4 | | 5 | Feb 2026 | huggingface.co | Official leaderboard | |
| 5 | GPT-6 Astra | 5 | Sep 2026 | developers.openai.com | Official leaderboard | |
| 6 | Claude Opus 5.5 | 5 | Sep 2026 | openrouter.ai | Official leaderboard | |
| 7 | GPT-6 Luna | 5 | Sep 2026 | developers.openai.com | Official leaderboard | |
| 8 | | 5 | Sep 2026 | cloud.google.com | Official leaderboard | |
| 9 | | 5 | Feb 2026 | openrouter.ai | Official leaderboard | |
| 10 | GPT-4o Aug 2024 | 5 | Aug 2024 | developers.openai.com | Official leaderboard |
Rankings by part
We read 5 parts from Slingshot AI’s leaderboard in 2 groups: clinical criteria (1-5) and safety flags. Each ranks models on one part of the test. GPT-4o leads 3 of them, and Gemini 2.5 Pro leads the other 2. Arrows show how far a model moved from its overall rank among the models we track.
Empathy
10 models we track have a score here.
- 1
Gemini 2.5 Pro Jun 2025 5
- 2
Qwen3.5-122B-A10B ↑2 5
- 3
Qwen3.5-27B ↑6 5
- 4 Claude Opus 5.5 ↑2 5
- 5
Gemini 4 Argon ↑3 5
- 6 GPT-5 ↓4 5
- 7 GPT-6 Astra ↓2 5
- 8 GPT-6 Luna ↓1 5
- 9 Claude Sonnet 4.5 ↓6 5
- 10 GPT-4o Aug 2024 5
Specificity
10 models we track have a score here.
- 1
Gemini 2.5 Pro Jun 2025 5
- 2 GPT-5 5
- 3 Claude Sonnet 4.5 5
- 4 GPT-6 Astra ↑1 5
- 5 Claude Opus 5.5 ↑1 5
- 6 GPT-6 Luna ↑1 5
- 7
Gemini 4 Argon ↑1 5
- 8
Qwen3.5-122B-A10B ↓4 5
- 9
Qwen3.5-27B 5
- 10 GPT-4o Aug 2024 5
Factual consistency
10 models we track have a score here.
- 1 GPT-4o Aug 2024 ↑9 4
- 2
Gemini 2.5 Pro Jun 2025 ↓1 4
- 3 GPT-5 ↓1 4
- 4 Claude Sonnet 4.5 ↓1 4
- 5
Qwen3.5-122B-A10B ↓1 4
- 6
Qwen3.5-27B ↑3 4
- 7 GPT-6 Astra ↓2 4
- 8 Claude Opus 5.5 ↓2 4
- 9 GPT-6 Luna ↓2 4
- 10
Gemini 4 Argon ↓2 4
Toxicity (1 to 5)
10 models we track have a score here.
- 1 GPT-4o Aug 2024 ↑9 1
- 2
Gemini 2.5 Pro Jun 2025 ↓1 1
- 3 GPT-5 ↓1 1
- 4 Claude Sonnet 4.5 ↓1 1
- 5
Qwen3.5-122B-A10B ↓1 1
- 6
Qwen3.5-27B ↑3 1
- 7 GPT-6 Astra ↓2 1
- 8 Claude Opus 5.5 ↓2 1
- 9 GPT-6 Luna ↓2 1
- 10
Gemini 4 Argon ↓2 1
Unauthorized medical advice (rate)
Fraction of answers flagged. 10 models we track have a score here.
- 1 GPT-4o Aug 2024 ↑9 0.16
- 2
Qwen3.5-122B-A10B ↑2 0.20
- 3 Claude Opus 5.5 ↑3 0.22
- 4
Qwen3.5-27B ↑5 0.24
- 5 GPT-6 Astra 0.27
- 6 GPT-6 Luna ↑1 0.27
- 7
Gemini 2.5 Pro Jun 2025 ↓6 0.31
- 8 Claude Sonnet 4.5 ↓5 0.31
- 9 GPT-5 ↓7 0.37
- 10
Gemini 4 Argon ↓2 0.38
2 models have held number one on Slingshot AI’s leaderboard since its release in May 2025
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Slingshot AI’s leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Slingshot AI’s 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 Law, finance and medicine
- MedXpertQA-MM In use Models answer expert-level medical exam questions that combine images with clinical records and test results. Scoring is accuracy.
Qwen 3.8 Max 80.4%
- HealthBench Professional In use 525 physician-authored tasks from real clinician chats (care consult, documentation, research) with physician-adjudicated rubrics. Claude Sonnet 5.5 69.2%
- HealthBench In use 5,000 realistic health conversations graded against physician-written rubrics. A length-adjusted variant penalizes verbosity. GPT-5 67.2%
- FinSearchComp In use Agents search the web to fetch live financial data, look up history, and run multi-step investigations, across 635 expert-written questions. Scored by answer accuracy.
Ling-3.1-flash 81.7%
- KorMedMCQA In use Models answer 7,469 multiple-choice questions from Korean licensing exams for doctors, nurses, pharmacists and dentists, held from 2012 to 2024. o1-preview 92.7%
- Harvey LAB-AA v1.1 In use Artificial Analysis's run of Harvey's Legal Agent Benchmark on Harvey's newer private set of 120 tasks across 24 practice areas. A deliverable with a material hallucination scores zero, and the headline is the share of judges who find every rubric criterion passed. GPT-6.2 Astra 15%
Data audit
All 10 checks pass. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 4 pass 1 note
- Pass Independent runs Scores from the official leaderboard (10 entries).
- Pass Models scored 10 models we track.
- Pass Newest score The newest is from September 2026, for Gemini 4 Argon.
- Pass Frontier models 3 of the top 10 have a score, led by GPT-6 Astra (#4).
- Note Headroom Scored as a rating, with no fixed ceiling.
Leaderboard import 3 pass
- Pass Importer We read the board on Slingshot AI: 10 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.