Vals MedScribe (MedScribe)

We show this table for reference; we do not rank on it.

Data verified October 29, 2026

Vals AI healthcare benchmark for whether models can support doctors with administrative work.

Evaluation results published by Vals AI. Published leaderboard aggregates and evaluation settings only. Dataset contents are not included.

MedScribe score on MedScribe — October 29, 2026

We mirror the published medscribe score view for MedScribe. Claude Opus 5.5 leads the public snapshot at 91.43%, followed by Claude Fable 5.1 (91.29%) and Claude Sonnet 5.5 (91.10%). We do not use these results to rank models overall.

1Closed

Claude Opus 5.5

Anthropic

91.43%

Overall 86.62Context 1M

2Closed

Claude Fable 5.1

Anthropic

91.29%

Overall 79.21Context 1M

3Closed

Claude Sonnet 5.5

Anthropic

91.10%

Overall 83.05Context 1M

109 modelsKnowledgeCurrentDisplay onlyUpdated October 29, 2026

MedScribe score table (109 models)

MedScribe score results for MedScribe
Rank	Model / configuration	Score	Parameters (B)	Open / closed
1	Claude Opus 5.5Anthropic	91.43%	Not reported	Closed
2	Claude Fable 5.1Anthropic	91.29%	Not reported	Closed
3	Claude Sonnet 5.5Anthropic	91.10%	Not reported	Closed
4	Claude Opus 5Anthropic	90.98%	Not reported	Closed
5	Muse Spark 1.2Metaxhigh reasoning	90.06%	Not reported	Closed
6	Claude Haiku 5.5Anthropic	89.54%	Not reported	Closed
7	Grok 4.7xAIxhigh reasoning	89.38%	Not reported	Closed
8	Muse Wonder FullMetamax reasoning	89.33%	Not reported	Closed
9	GLM-5.3-FlashZ.AImax reasoning	88.94%	Not reported	Open
10	Muse Spark 1.1Metaxhigh reasoning	88.89%	Not reported	Closed
11	GLM-5.3Z.AImax reasoning	88.81%	Not reported	Open
12	Claude Fable 5Anthropic	88.52%	Not reported	Closed
13	MiMo-V2.6-ProXiaomi	88.31%	Not reported	Open
14	GPT-5.1OpenAIhigh reasoning	88.09%	Not reported	Closed
15	Kimi K3Moonshot AImax reasoning	88.05%	Not reported	Open
16	GPT-6 AstraOpenAImax reasoning	87.91%	Not reported	Closed
17	Gemini 4 ArgonGooglehigh reasoning	87.43%	Not reported	Closed
18	MiniMax M3MiniMax	87.25%	Not reported	Open
19	Grok 4.5xAIhigh reasoning	86.88%	Not reported	Closed
20	GPT-5.5OpenAIxhigh reasoning	86.87%	Not reported	Closed
21	Claude Opus 4.6Anthropic	86.74%	Not reported	Closed
22	Grok 4.6xAIhigh reasoning	86.53%	Not reported	Closed
23	GPT-6.1 SolOpenAImax reasoning	86.45%	Not reported	Closed
24	Claude Opus 4.6 (Adaptive)Anthropic	86.13%	Not reported	Closed
25	Muse SparkMeta	85.90%	Not reported	Closed
26	Claude Opus 4.8Anthropic	85.75%	Not reported	Closed
27	DeepSeek V4.1 Flash 1017DeepSeekhigh reasoning	85.50%	Not reported	Open
28	InklingThinking Machines Lab0.99 reasoning	85.41%	Not reported	Open
29	Claude Opus 4.5 ThinkingAnthropic	85.32%	Not reported	Closed
30	MiMo-V2.6-FlashXiaomi	85.28%	Not reported	Open
31	GPT-5.6 SolOpenAImax reasoning	85.23%	Not reported	Closed
32	Claude Haiku 4.5 ThinkingAnthropic	85.23%	Not reported	Closed
33	Qwen3.8 MaxAlibaba	84.95%	Not reported	Open
34	Claude Sonnet 4.5Anthropic	84.52%	Not reported	Closed
35	Gemini 3.8 FlashGooglehigh reasoning	84.50%	Not reported	Closed
36	GPT-5.6 LunaOpenAImax reasoning	84.39%	Not reported	Closed
37	GPT-5.2OpenAIxhigh reasoning	84.39%	Not reported	Closed
38	Inkling-SmallThinking Machines Lab0.99 reasoning	84.11%	Not reported	Open
39	Claude Sonnet 4.5 ThinkingAnthropic	84.10%	Not reported	Closed
40	Gemini 3.7 FlashGooglehigh reasoning	83.94%	Not reported	Closed
41	Qwen3.8-27BAlibabaxhigh reasoning	83.85%	Not reported	Open
42	MiMo-V2.5-ProXiaomi	83.73%	Not reported	Closed
43	GPT-6 LunaOpenAImax reasoning	83.71%	Not reported	Closed
44	GPT-5OpenAIhigh reasoning	83.65%	Not reported	Unknown
45	Hy4 previewTencent	83.60%	Not reported	Open
46	GLM-5.2Z.AI	83.53%	Not reported	Open
47	Claude Opus 4.5Anthropic	83.25%	Not reported	Closed
48	Gemini 2.5 Flash ThinkingGoogle	82.98%	Not reported	Unknown
49	Claude Opus 4.7Anthropic	82.95%	Not reported	Closed
50	Gemini 2.5 FlashGoogle	82.87%	Not reported	Closed
51	GPT-5.6 TerraOpenAIxhigh reasoning	82.87%	Not reported	Closed
52	GPT-6 SolOpenAImax reasoning	82.03%	Not reported	Closed
53	Grok 4 Fast (Reasoning)xAI	81.63%	Not reported	Closed
54	Ling 3.0 Flash 2607Ant	80.90%	Not reported	Unknown
55	MiniMax M2.1MiniMax	80.78%	Not reported	Unknown
56	GPT-5 miniOpenAIhigh reasoning	80.58%	Not reported	Closed
57	Mistral Large 4Mistralhigh reasoning	80.43%	Not reported	Open
58	DeepSeek V4 Flash 0731DeepSeekhigh reasoning	80.36%	Not reported	Open
59	DeepSeek V4 Pro 0813DeepSeekmax reasoning	80.17%	Not reported	Open
60	MiniMax M2.7MiniMax	79.87%	Not reported	Open
61	Grok 4 Fast Non ReasoningxAI	79.72%	Not reported	Unknown
62	Gemini 3.6 FlashGooglehigh reasoning	79.66%	Not reported	Closed
63	Qwen3.7 MaxAlibaba	79.40%	Not reported	Closed
64	Grok 4.1 Fast (Reasoning)xAI	78.73%	Not reported	Closed
65	Gemini 2.5 Flash Preview 09 2025 ThinkingGoogle	78.50%	Not reported	Unknown
66	Grok 4 0709xAI	78.15%	Not reported	Unknown
67	Kimi K2.6Moonshot AI	78.15%	Not reported	Open
68	Gemini 2.5 Flash Preview 09 2025Google	77.95%	Not reported	Unknown
69	GPT-5.4OpenAIxhigh reasoning	77.55%	Not reported	Closed
70	Grok 4.1 Fast Non ReasoningxAI	77.46%	Not reported	Unknown
71	Qwen3 VL PlusAlibaba	77.13%	Not reported	Unknown
72	GPT-5.4 nanoOpenAIhigh reasoning	77.09%	Not reported	Closed
73	Qwen3.6 PlusAlibaba	76.96%	Not reported	Closed
74	o3OpenAIhigh reasoning	76.65%	Not reported	Closed
75	Gemini 3.5 FlashGooglehigh reasoning	76.57%	Not reported	Closed
76	Kimi K2.5 ThinkingMoonshot AI	76.44%	Not reported	Unknown
77	Gemini 3.1 Pro PreviewGooglehigh reasoning	76.11%	Not reported	Unknown
78	Claude Sonnet 5Anthropic	76.05%	Not reported	Closed
79	Gemini 2.5 Flash Lite Preview 09 2025Google	75.82%	Not reported	Unknown
80	Ling 3.0 Flash Af Rc3Ant	75.59%	Not reported	Unknown
81	DeepSeek V4 Pro 0813DeepSeekmax reasoning	75.14%	Not reported	Open
82	Grok 4.3xAI	74.40%	Not reported	Closed
83	Claude Opus 4.1 20250805 ThinkingAnthropic	73.90%	Not reported	Unknown
84	Gemini 2.5 ProGoogle	73.55%	Not reported	Closed
85	GPT-5 nanoOpenAIhigh reasoning	72.86%	Not reported	Closed
86	Gemini 2.5 Flash-LiteGoogle	72.83%	Not reported	Closed
87	Qwen3 MaxAlibaba	72.71%	Not reported	Closed
88	Claude Sonnet 4Anthropic	72.41%	Not reported	Unknown
89	GLM-5.1Z.AI	72.27%	Not reported	Open
90	MiMo-V2.5Xiaomi	72.15%	Not reported	Closed
91	Gemini 3 Pro PreviewGooglehigh reasoning	72.04%	Not reported	Unknown
92	Claude Opus 4.1Anthropic	71.75%	Not reported	Unknown
93	Gemini 3.5 Flash-LiteGooglehigh reasoning	70.89%	Not reported	Closed
94	Qwen3.5 FlashAlibaba	70.62%	Not reported	Closed
95	Gemini 3 Flash PreviewGooglehigh reasoning	69.92%	Not reported	Unknown
96	Claude Sonnet 4 20250514 ThinkingAnthropic	69.35%	Not reported	Unknown
97	o4-miniOpenAIhigh reasoning	69.14%	Not reported	Closed
98	GLM-4.7Z.AI	68.63%	Not reported	Open
99	Mistral Medium 3.5Mistral AIhigh reasoning	67.73%	Not reported	Unknown
100	Gemini 2.5 Flash Lite Preview 09 2025 ThinkingGoogle	66.88%	Not reported	Unknown
101	Laguna M.1Poolside	65.91%	Not reported	Closed
102	Gemini 3.1 Flash Lite PreviewGooglehigh reasoning	63.90%	Not reported	Unknown
103	Grok 4.20 0309 ReasoningxAI	63.41%	Not reported	Unknown
104	Laguna XS.2Poolside	61.43%	Not reported	Open
105	Command A Plus 05 2026Cohere	55.68%	Not reported	Unknown
106	Mercury 2.5Inceptionhigh reasoning	55.09%	Not reported	Closed
107	Llama4 Maverick Instruct BasicFireworks AI	54.22%	Not reported	Unknown
108	Meta Llama Llama 4 Scout 17B 16E InstructTogether AI	50.59%	Not reported	Unknown
109	Nemotron Lightning 3p5 30b A3bFireworks AI	4.27%	Not reported	Unknown

How MedScribe is shown here

BenchLM mirrors the public Vals AI MedScribe leaderboard captured from https://www.vals.ai/benchmarks/medscribe and updated by Vals on October 29, 2026. The snapshot preserves overall scores, uncertainty, latency, cost-per-test metadata, and task-level scores where Vals publishes them.

MedScribe is display only on BenchLM. Vals proprietary or Vals-hosted aggregate views are useful context, but BenchLM does not use them as weighted ranking inputs or as a replacement for benchmark-native source records.

Snapshot

109 Vals rows1 task viewsprivate datasetTasks: OverallDisplay only

MedScribe on Vals AI Vals methodology Vals home

The published MedScribe snapshot places Claude Opus 5.5 first at 91.43%. The third row is 0.33 points behind. The broader top-10 range is 2.54 points, so many of the published results sit in a relatively narrow band.

109 models have been evaluated on MedScribe. The benchmark falls in the Knowledge category. We keep external benchmark mirrors separate from the weighted global scoring system, so these results remain source-specific evidence. MedScribe is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.

About MedScribe

Year

2026

Tasks

Medical administrative support tasks

Format

Accuracy score

Difficulty

Professional healthcare administration

BenchLM mirrors the public Vals MedScribe leaderboard as display-only healthcare evidence.

MedScribe Public benchmark source

Freshness and provenance

Version

MedScribe 2026

Refresh cadence

Quarterly

Staleness state

Current

Question availability

Public benchmark set

CurrentDisplay only

BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.

Questions

What does MedScribe measure?

Vals AI healthcare benchmark for whether models can support doctors with administrative work.

Which model leads the published MedScribe snapshot?

Claude Opus 5.5 currently leads the published MedScribe snapshot with 91.43% medscribe score. BenchLM shows this benchmark for display only and does not use it in overall rankings.

How many models are evaluated on MedScribe?

The October 29, 2026 snapshot contains 109 AI models.

Last updated: October 29, 2026 · mirrored from the public benchmark leaderboard

Know when it’s worth switching models

The model to choose, the cheaper alternative, and the release we would wait on.

Read a sample issue

Join 7,000+ subscribers.

One email each week. Unsubscribe anytime.