Benchmarks › Vals MedScribe
91.4top score · Claude Opus 5.5
80models ranked
29open-weights models
SOAP rubric-check accuracyhigher is better
Oct 20, 2026last updated
How the test works
1
Task
100 transcript rubrics
- What the model doesWrite clinical SOAP notes from encounter information
- Kinds of tasksSOAP note generation
2
Run
Shared expert-defined SOAP template
- Token budget30,000 generated tokens maximum
- Temperature1
3
Score
SOAP rubric-check accuracy
- Graded byLLM judge checks generated notes against expert-consensus SOAP rubrics
- Better isHigher
Sources: vals.ai
Vals MedScribe leaderboardtop 10 of 80
Full leaderboard with filters →| Model | Vals MedScribe | Per task | Price | Speed | Supports | Context |
|---|---|---|---|---|---|---|
91.4unstated | — | $4 / $20 | — | 1M | ||
91.3unstated | — | $10 / $50 | — | 1M | ||
91.1unstated | — | $2 / $10 | — | 1M | ||
91unstated | — | $5 / $25 | — | 1M | ||
90.1unstated | — | $1.25 / $4.25 | — | 1M | ||
89.5unstated | — | $0.10 / $0.50 | — | 1M | ||
89.4unstated | — | $2 / $6 | — | 500K | ||
88.9unstated | — | $0.15 / $0.50 | — | 1M | ||
88.9unstated | — | $1.25 / $4.25 | — | 1M | ||
88.8unstated | — | $1.40 / $4.40 | — | 1M |
Vals MedScribe questions
What is Vals MedScribe?›
Write clinical SOAP notes from encounter information. It has 100 transcript rubrics. Created by Vals AI and Protege.
Which model is best on Vals MedScribe?›
Claude Opus 5.5 leads Vals MedScribe with 91.4, ahead of Claude Fable 5.1 (91.3) and Claude Sonnet 5.5 (91.1).
What is the best open-weights model on Vals MedScribe?›
GLM 5.3 Flash with 88.9, then GLM 5.3 (88.8).
How is Vals MedScribe scored?›
SOAP rubric-check accuracy; llm judge checks generated notes against expert-consensus soap rubrics.