MLCR-AA: Medical Long Context Reasoning

Un benchmark ouvert de Wisedocs qui mesure la capacité des modèles à raisonner sur des dossiers médicaux longs et fragmentés, en réalisant la synthèse multi-documents sur laquelle s'appuient les professionnels des demandes lors de l'examen de dossiers d'assurance et de santé.

Modèles: 96 actuel, 181 au total
Médiane: 15.8%
Plage: 0.0%–43.9%

Distribution des scores

Le plus bas: 0.0%
Médiane: 15.8%
Le plus élevé: 43.9%

Classement

Inclure les modèles obsolètes

Rang	Modèle	Score
1	Claude Fable 5.2 (Max, Default Fallback) Anthropic	43.9%
2	Claude Sonnet 5.5 (Max, Default Fallback) Anthropic	41.7%
3	GPT-6 Astra (Max) OpenAI	41.7%
4	GPT-6.1 Sol (Max) OpenAI	38.3%
5	GPT-6 Astra (Medium) OpenAI	38.3%
6	Kimi K3.1 Pro Kimi	38.3%
7	GPT-6 Astra (Xhigh) OpenAI	36.1%
8	GPT-6.1 Sol (Xhigh) OpenAI	35.6%
9	Claude Haiku 5.5 (Max) Anthropic	35.0%
10	GPT-6 Astra (Low) OpenAI	33.9%
11	GPT-6 Astra (High) OpenAI	32.8%
12	GPT-6.1 Sol (Medium) OpenAI	32.2%
13	GPT-5.5 Instant (June 2026) OpenAI	31.7%
14	GPT-6.1 Sol (High) OpenAI	31.7%
15	GPT-6.1 Terra (Max) OpenAI	30.6%
16	Qwen3.8 2.4T A95B Alibaba	30.6%
17	Claude Opus 5.5 (Max, Default Fallback) Anthropic	27.8%
18	DeepSeek V4.1 Flash 1017 (Max) DeepSeek	26.7%
19	GPT-6.1 Sol (Low) OpenAI	26.7%
20	GPT-6 Luna (Max) OpenAI	26.1%
21	Claude Opus 5.5 (Xhigh, Default Fallback) Anthropic	25.0%
22	Qwen3.8 27B (Xhigh) Alibaba	22.8%
23	Claude Opus 5.5 (Medium, Default Fallback) Anthropic	22.2%
24	Muse Wonder (Max) Meta	22.2%
25	Muse Wonder Full (Max) Meta	22.2%
26	Grok 4.7 (Low) SpaceXAI	21.1%
27	Claude Sonnet 5.5 (Xhigh, Default Fallback) Anthropic	20.6%
28	GPT-6 Luna (Xhigh) OpenAI	20.6%
29	Inkling 2 Small Thinking Machines	20.6%
30	Qwen3.8-Flash-Next Alibaba	20.6%
31	Grok 4.7 (Xhigh) SpaceXAI	20.0%
32	Qwen3.8 Max (0902) Alibaba	20.0%
33	Claude Opus 5.5 (Low, Default Fallback) Anthropic	19.4%
34	Claude Sonnet 5.5 (Low, Default Fallback) Anthropic	19.4%
35	Gemini 4 Carbon (High) Google	19.4%
36	GPT-6 Luna (High) OpenAI	19.4%
37	Claude Sonnet 5.5 (High, Default Fallback) Anthropic	18.9%
38	GPT-6 Luna (Low) OpenAI	18.9%
39	Grok 4.7 (High) SpaceXAI	18.9%
40	GPT-6 Luna (Medium) OpenAI	18.3%
41	Muse Ember (Max) Meta	18.3%
42	Step 5 Preview StepFun	18.3%
43	Claude Sonnet 5.5 (Medium, Default Fallback) Anthropic	17.8%
44	Mistral Large 4 (High) Mistral	17.8%
45	Beam 1.0 Reflection	17.8%
46	GLM-5.3 (Max) Z AI	17.2%
47	MiniMax-M3.1-Flash MiniMax	16.7%
48	Claude Haiku 5.5 (Medium) Anthropic	16.1%
49	Gemini 3.8 Flash (Low) Google	15.6%
50	Claude Haiku 5.5 (High) Anthropic	15.0%