Skip to content

Benchmark

GDM-MRCRv2: 8 needle 256k

Long context benchmarks measure how well models use information in long inputs. Scores are percentages.

Models measured
45
Scores recorded
128
Sources
1
IQ weight
1.85

Score distribution

39.4%
Mean
39.8%
Median
21%
Std dev
10.4%
10th percentile
69.1%
90th percentile
1.2%
Min
1.2%89%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.85 for chat models, measured over 45 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
51GPT-5.6 Lunamax· Context Arena data45.1%
52Gemini 3 Flashhigh· Context Arena data44.9%
53Qwen3.7 Max· Context Arena data44.5%
54DeepSeek V4 Flash 0731max· Context Arena data44.5%
55Gemini 3.1 Pro Previewlow· Context Arena data44.1%
56GPT 5.4low· Context Arena data43.4%
57Qwen3.7 Flashenabled· Context Arena data43.1%
58Grok 4.6low· Context Arena data41.8%
59Gemini 3.5 Flashmedium· Context Arena data41.7%
60Qwen3.6 Plus· Context Arena data41%
61Gemini 3.5 Flash Litemedium· Context Arena data40.3%
62Qwen3.7 Flash· Context Arena data40.1%
63Claude Opus 4.6· Context Arena data40%
64Gemini 3 Flashmedium· Context Arena data39.8%
65Gemini 3 Flashlow· Context Arena data39.6%
66DeepSeek V4 Flash 0731high· Context Arena data39.6%
67Qwen3.6 Plusenabled· Context Arena data39%
68DeepSeek V4 Pro 0813high· Context Arena data38.9%
69Qwen3.6 Flash· Context Arena data38.1%
70Gemini 3.5 Flash Litehigh· Context Arena data37.7%
71DeepSeek V4 Pro 0813max· Context Arena data36.1%
72Gemini 3.5 Flash Litelow· Context Arena data35.8%
73Kimi K3max· Context Arena data34.8%
74Gemini 3.6 Flashminimal· Context Arena data34.2%
75DeepSeek V4 Pro 0813low· Context Arena data34.1%
76DeepSeek V4 Proenabled· Context Arena data33.2%
77GPT 5.4 Nanoxhigh· Context Arena data33%
78DeepSeek V4 Flashenabled· Context Arena data31.3%
79Qwen3.5 Plusenabled· Context Arena data29.7%
80DeepSeek V4 Flash 0731low· Context Arena data29.5%
81Kimi K3high· Context Arena data29.4%
82GPT-5.4 minihigh· Context Arena data28.3%
83MiMo V2.5enabled· Context Arena data28%
84Gemini 3 Flash· Context Arena data27.7%
85GPT 5.5· Context Arena data27.5%
86Gemini 3.5 Flashlow· Context Arena data27.5%
87Qwen3.5 Plus· Context Arena data26.4%
88GPT 5.4 Nanohigh· Context Arena data26.3%
89DeepSeek V4 Flash 0731· Context Arena data26%
90DeepSeek V4 Pro 0813· Context Arena data25.9%
91GPT-5.4 minixhigh· Context Arena data25.2%
92Gemini 3.5 Flash· Context Arena data24.7%
93MiMo V2 Pro· Context Arena data23.5%
94MiMo V2.5· Context Arena data23.5%
95GPT-5.4 minimedium· Context Arena data23%
96DeepSeek V4 Pro· Context Arena data22.6%
97Gemini 3.1 Flash Litehigh· Context Arena data22.5%
98DeepSeek V4 Flash· Context Arena data22.4%
99Gemini 3.1 Flash Litelow· Context Arena data22.1%
100Kimi K3low· Context Arena data21.8%

Showing 51–100 of 128 results.