Long context

MLCR-AA leaderboard

Name: benchgap: LLM benchmark scores, measured and estimated
Creator: benchgap
License: https://creativecommons.org/licenses/by-nc/4.0/

As of 2026-11-05, the highest measured score on MLCR-AA is 41.7% by Claude Sonnet 5.5. 106 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	Claude Sonnet 5.5	41.7%	measured
2	GPT-6 Astra	41.7%	measured
3	GPT-6.1 Sol	38.3%	measured
4	Kimi K3.1 Pro	38.3%	measured
5	Claude Haiku 5.5	35.0%	measured
6	GPT-6 Sol	33.9%	measured
7	GPT-6.1 Terra	30.6%	measured
8	Kimi K3	28.2%	estimated ± 10.5 pp, low confidence
9	Step 5 Preview	28.2%	estimated ± 10.5 pp, low confidence
10	GPT-5.5	28.2%	estimated ± 10.5 pp, low confidence
11	DeepSeek V4.1 Flash 1017	28.2%	estimated ± 10.5 pp, low confidence
12	GPT-5.6 Sol	28.2%	estimated ± 10.5 pp, low confidence
13	GPT-5.6 Luna	28.2%	estimated ± 10.5 pp, low confidence
14	GPT-5.3 Codex	28.2%	estimated ± 10.5 pp, low confidence
15	Muse Glimmer 30B	28.2%	estimated ± 10.5 pp, low confidence
16	GPT-5.6 Terra	28.2%	estimated ± 10.5 pp, low confidence
17	Ling 3.1 Flash	28.2%	estimated ± 10.5 pp, low confidence
18	MiniMax M3	28.2%	estimated ± 10.5 pp, low confidence
19	Muse Spark 1.3	28.2%	estimated ± 10.5 pp, low confidence
20	GPT-5.2	28.2%	estimated ± 10.5 pp, low confidence
21	Claude Fable 5	28.1%	estimated ± 10.5 pp, low confidence
22	GPT-5.2-Codex	28.1%	estimated ± 10.5 pp, low confidence
23	Claude Sonnet 5	28.0%	estimated ± 10.5 pp, low confidence
24	Gemini 3.1 Pro	28.0%	estimated ± 10.5 pp, low confidence
25	GPT-5.4	28.0%	estimated ± 10.5 pp, low confidence
26	Gemini 3.7 Flash	27.8%	estimated ± 10.5 pp, low confidence
27	Claude Opus 5.5	27.8%	measured
28	Kimi K2.6	26.6%	estimated ± 10.5 pp, low confidence
29	Sage-A1.5	26.6%	estimated ± 10.5 pp, low confidence
30	GPT-6 Luna	26.1%	measured
31	Qwen 3.6 Max (preview)	25.6%	estimated ± 10.5 pp, low confidence
32	DeepSeek V4 Pro 0813	23.3%	estimated ± 10.5 pp, low confidence
33	Grok 4.6	23.3%	estimated ± 10.5 pp, low confidence
34	Qwen3.8 Max Preview	23.3%	estimated ± 10.5 pp, low confidence
35	Qwen3.8-27B	22.8%	measured
36	Muse Wonder	22.2%	measured
37	Muse Wonder Full	22.2%	measured
38	Gemini 3.6 Flash	21.1%	estimated ± 10.5 pp, low confidence
39	GPT-5.1	21.1%	estimated ± 10.5 pp, low confidence
40	Grok 4.7	20.0%	measured
41	DeepSeek V4 Flash 0731	18.7%	estimated ± 10.5 pp, low confidence
42	Gemini 4 Argon	18.7%	estimated ± 10.5 pp, low confidence
43	MiMo-V2.5-Pro	18.7%	estimated ± 10.5 pp, low confidence
44	Qwen3.8-Flash-Next	18.7%	estimated ± 10.5 pp, low confidence
45	Beam 1.0 501B-A23B	17.8%	measured
46	GLM-5.3	17.2%	measured
47	MiniMax M3.1 Flash	16.7%	measured
48	Beam 1.0 501B-A23B FP8	16.6%	estimated ± 10.5 pp, low confidence
49	Apodex 1.1	16.0%	estimated ± 10.5 pp, low confidence
50	Apodex 1.1 Mini	16.0%	estimated ± 10.5 pp, low confidence
51	Claude Opus 5	16.0%	estimated ± 10.5 pp, low confidence
52	Gemini 4 Carbon	16.0%	estimated ± 10.5 pp, low confidence
53	Grok 4.5	16.0%	estimated ± 10.5 pp, low confidence
54	Kimi K2.7 Code	16.0%	estimated ± 10.5 pp, low confidence
55	Claude Fable 5.1	15.6%	measured
56	Hy3	14.6%	estimated ± 10.5 pp, low confidence
57	Muse Spark 1.2	14.6%	estimated ± 10.5 pp, low confidence
58	Qwen3.7 Max	14.6%	estimated ± 10.5 pp, low confidence
59	Claude Opus 4.7 (Adaptive)	13.7%	estimated ± 10.5 pp, low confidence
60	Beam 1.0 501B-A23B NVFP4	13.7%	estimated ± 10.5 pp, low confidence
61	GLM-5.2	13.0%	estimated ± 10.5 pp, low confidence
62	Ling 3.0 Flash VL	13.0%	estimated ± 10.5 pp, low confidence
63	MiniMax M2.7	13.0%	estimated ± 10.5 pp, low confidence
64	Qwen3.6 Plus	13.0%	estimated ± 10.5 pp, low confidence
65	GPT-5 (high)	12.9%	estimated ± 10.5 pp, low confidence
66	MiMo-V2.6-Pro	12.8%	measured
67	Claude Opus 4.6 (Adaptive)	12.8%	estimated ± 10.5 pp, low confidence
68	Kimi K2.5	12.8%	estimated ± 10.5 pp, low confidence
69	Kimi K2.5 (Reasoning)	12.8%	estimated ± 10.5 pp, low confidence
70	Muse Spark	12.8%	estimated ± 10.5 pp, low confidence
71	Claude Opus 4.8	12.6%	estimated ± 10.5 pp, low confidence
72	Muse Spark 1.1	12.6%	estimated ± 10.5 pp, low confidence
73	Qwen3.5-27B	12.6%	estimated ± 10.5 pp, low confidence
74	Claude Opus 4.5 Thinking	12.5%	estimated ± 10.5 pp, low confidence
75	Qwen3.5 397B	12.5%	estimated ± 10.5 pp, low confidence
76	Qwen3.5 397B (Reasoning)	12.5%	estimated ± 10.5 pp, low confidence
77	Qwen3.6-27B	12.5%	estimated ± 10.5 pp, low confidence
78	GPT-5.4 mini	12.5%	estimated ± 10.5 pp, low confidence
79	Ternary Bonsai 2 27B	12.5%	estimated ± 10.5 pp, low confidence
80	GPT-5.4 nano	12.4%	estimated ± 10.5 pp, low confidence
81	Quasar 438B	12.4%	estimated ± 10.5 pp, low confidence
82	Qwen3.5-122B-A10B	12.4%	estimated ± 10.5 pp, low confidence
83	Claude 4.1 Opus Thinking	12.4%	estimated ± 10.5 pp, low confidence
84	Gemini 3.5 Flash-Lite	12.4%	estimated ± 10.5 pp, low confidence
85	Gemini 3 Pro	12.4%	estimated ± 10.5 pp, low confidence
86	GPT-5 (medium)	12.4%	estimated ± 10.5 pp, low confidence
87	Claude Opus 4.7	12.4%	estimated ± 10.5 pp, low confidence
88	GLM-5	12.4%	estimated ± 10.5 pp, low confidence
89	Inkling-Small	12.4%	estimated ± 10.5 pp, low confidence
90	MiMo-V2-Omni	12.4%	estimated ± 10.5 pp, low confidence
91	Ripple-65B Preview	12.4%	estimated ± 10.5 pp, low confidence
92	o3	12.4%	estimated ± 10.5 pp, low confidence
93	MiMo-V2.6-Flash	12.4%	estimated ± 10.5 pp, low confidence
94	Grok 4.1 Fast (Reasoning)	12.4%	estimated ± 10.5 pp, low confidence
95	GLM-5.1	12.4%	estimated ± 10.5 pp, low confidence
96	Grok 4 Fast (Reasoning)	12.4%	estimated ± 10.5 pp, low confidence
97	Step 3.7 Flash	12.4%	estimated ± 10.5 pp, low confidence
98	Ling 3.0 Flash	12.4%	estimated ± 10.5 pp, low confidence
99	Qwen3.7 Plus	12.4%	estimated ± 10.5 pp, low confidence
100	Qwen3.5-35B-A3B	12.4%	estimated ± 10.5 pp, low confidence
101	GLM-5-Turbo	12.4%	estimated ± 10.5 pp, low confidence
102	Qwen3.6-35B-A3B	12.4%	estimated ± 10.5 pp, low confidence
103	GLM-4.7	12.4%	estimated ± 10.5 pp, low confidence
104	Solar Pro 4	12.4%	estimated ± 10.5 pp, low confidence
105	Claude Opus 4.5	12.4%	estimated ± 10.5 pp, low confidence
106	Ling 3.0 Flash Sante	12.4%	estimated ± 10.5 pp, low confidence
107	GLM-5V-Turbo	12.4%	estimated ± 10.5 pp, low confidence
108	Gemma 4 31B	12.4%	estimated ± 10.5 pp, low confidence
109	Nemotron 3.5 Bolt 74B-A9B	12.4%	estimated ± 10.5 pp, low confidence
110	Claude 4 Opus Thinking	12.4%	estimated ± 10.5 pp, low confidence
111	Gemini 3.5 Flash	12.4%	estimated ± 10.5 pp, low confidence
112	GPT-5.1-Codex	12.4%	estimated ± 10.5 pp, low confidence
113	GPT-5.1-Codex-Max	12.4%	estimated ± 10.5 pp, low confidence
114	Mistral Medium 3.5 128B	12.4%	estimated ± 10.5 pp, low confidence
115	Gemini 2.5 Pro	12.4%	estimated ± 10.5 pp, low confidence
116	Nemotron 3.5 Super VL 120B A12B	12.4%	estimated ± 10.5 pp, low confidence
117	Claude Sonnet 4.6	12.4%	estimated ± 10.5 pp, low confidence
118	GPT-4.1	12.4%	estimated ± 10.5 pp, low confidence
119	MiMo-V2-Pro	12.4%	estimated ± 10.5 pp, low confidence
120	Grok 4	12.4%	estimated ± 10.5 pp, low confidence
121	Mercury 2.5	12.4%	estimated ± 10.5 pp, low confidence
122	Nemotron 3.5 Super VL 120B A12B NVFP4	12.4%	estimated ± 10.5 pp, low confidence
123	Nemotron 3.5 Bolt 74B-A9B BF16	12.4%	estimated ± 10.5 pp, low confidence
124	Spark-X2.6-4B	12.4%	estimated ± 10.5 pp, low confidence
125	Gemini 3.8 Flash	10.6%	measured
126	Muse Spark Open	7.8%	measured
127	Inkling	6.1%	measured

MLCR-AA leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General