MMODELYST
Evals/AA Long-Context Reasoning
EVL

AA Long-Context Reasoning

benchmarkLong Contextaccuracy %
Leaderboard · 100 models
accuracy % — higher is better
1GPT-5.2 Codex
75.7
2GPT-5
75.6
3GPT-5.1
75
4GPT-5.5
74.3
5Claude Opus 4.5
74
6GPT-5.4
74
7KAT-Coder-Pro V1
74
8MiniMax-M3
74
9GPT-5.6 Luna (max)
74
10GPT-5.3 Codex
74
11GPT-5.6 Terra (max)
74
12GPT-5.6 Sol (max)
73.7
13MiMo-V2.5-Pro
73.3
14GPT-5.2
72.7
15Gemini 3.1 Pro Preview
72.7
16GLM-5.2 (max)
71.3
17GPT-5.6 Terra
71.3
18GPT-5.6 Sol
71
19Claude Opus 4.6
70.7
20Gemini 3 Pro Preview
70.7
21Claude Sonnet 4.6
70.7
22Claude Sonnet 5
70.7
23Claude Opus 4.7
70.3
24Claude 4.5 Haiku
70.3
25Claude Fable 5
70
26Qwen3.6 Max Preview
69.7
27Qwen3.6 Plus
69.7
28Kimi K2.6
69.7
29Muse Spark
69.7
30GPT-5.6 Luna
69.7
31Gemini 3.5 Flash
69.3
32o3
69.3
33GPT-5.4 mini
69.3
34DeepSeek V3.2 Exp
69
35GPT-5 Codex
69
36Qwen3.7 Max
69
37Qwen3.6 27B
68.7
38MiniMax-M2.7
68.7
39Grok 4
68
40Grok 4.1 Fast
68
41Nex-N2-Pro
67.7
42Grok 4.5
67.7
43Claude Opus 4.8
67.7
44Qwen3.5 27B
67.3
45GPT-5.1 Codex
67.3
46Qwen3 235B A22B 2507
67
47Nemotron 3 Ultra 550B A55B
67
48Qwen3.5 122B A10B
66.7
49MiMo-V2-Omni
66.7
50Kimi K2 Thinking
66.3
51Claude 4.1 Opus
66.3
52Gemini 3 Flash Preview
66.3
53Kimi K2.7 Code
66.3
54DeepSeek V4 Pro
66.3
55Qwen3 Max Thinking
66
56MiniMax-M2.5
66
57GPT-5 mini
66
58KAT Coder Pro V2
66
59Gemini 2.5 Pro
66
60GPT-5.4 nano
66
61Claude 4.5 Sonnet
65.7
62Qwen3.5 397B A17B
65.7
63Kimi K2.5
65.3
64Doubao Seed Code
65.3
65Gemini 3.1 Flash-Lite
65.3
66DeepSeek V3.1 Terminus
65
67DeepSeek V3.2
65
68Qwen3.7 Plus
65
69Grok 4 Fast
64.7
70Claude 4 Sonnet
64.7
71Grok Build 0.1 0616
64.7
72Gemini 2.5 Flash Preview (Sep '25)
64.3
73Ring-2.6-1T
64.3
74MiMo-V2-Flash (Feb 2026)
64.3
75Grok 4.3
64.3
76GLM-4.7
64
77GPT-5.5 Instant (May 2026)
64
78GPT-5 (ChatGPT)
63.7
79Qwen3.6 35B A3B
63.7
80MiMo-V2-Omni-0327
63.7
81Step 3.7 Flash
63.7
82GLM-5
63.3
83Muse Spark 1.1
63.3
84MiMo-V2-Flash
63
85DeepSeek V4 Flash
63
86GPT-5.1 Codex mini
62.7
87Qwen3.5 35B A3B
62.7
88MiMo-V2.5
62.7
89GLM-5.1
62.3
90Gemma 4 31B
62
91Gemini 2.5 Flash
61.7
92MiniMax-M2
61
93GPT-4.1
61
94GLM 5V Turbo
61
95Mistral Medium 3.5
61
96MiMo-V2-Pro
60.7
97Claude 3.7 Sonnet
60.7
98GLM-5-Turbo
60.7
99Qwen3 Next 80B A3B
60.3
100NVIDIA Nemotron 3 Super 120B A12B
60
Benchmark health
active

Discriminating but climbing — the frontier is moving up this benchmark.

74.5
top-10 mean
75.7
best
100
models
Recent changes
GLM-5.2+37.36d ago
JT-4.1 Flash 236B A21B+58.76d ago
Grok 4.5+67.76d ago
Muse Spark 1.1+63.36d ago
GPT-5.6 Luna (max)+746d ago
What improves this score

Datasets & environments shown to raise it.

No data yet.