Documentation and coding benchmarks

2 tracked · snapshot reviewed September 28, 2026

These benchmarks evaluate clinical documentation and medical coding, including note generation and assignment of diagnosis codes. Their task definitions, reference answers, and grading criteria differ, so each score should be read within its own board.

MedCode (Vals AI)

Vals AI · 2,755 patient records

ICD-10-CM coding of whole hospital stays from discharge summaries and notes, checked against professional coders.

Showing top 12 of 102 indexed results. View all results.

MedScribe (Vals AI)

Vals AI · 100 SOAP-note cases

Quality and compliance of SOAP notes generated from clinical visits, scored against documentation rubrics.

91.43%
Claude Opus 5.5
91.29%
Claude Fable 5.1
91.10%
Claude Sonnet 5.5
90.98%
Claude Opus 5
90.06%
Muse Spark 1.2
89.38%
S
Grok 4.7
88.94%
Z
GLM 5.3 Flash
88.89%
Muse Spark 1.1
88.81%
Z
GLM 5.3
88.52%
Claude Fable 5
88.31%
X
MiMo V2.6 Pro
88.09%
GPT 5.1

Showing top 12 of 104 indexed results. View all results.

Which documentation and coding benchmarks have results in this index?

2 as of September 28, 2026: MedCode (Vals AI) (Claude Opus 5: highest indexed score 63.57%); MedScribe (Vals AI) (Claude Opus 5.5: highest indexed score 91.43%).

The other categories sit on the index: rubric-graded benchmarks, agentic and workflow benchmarks, safety benchmarks, knowledge and exam benchmarks, composite indices.