Skip to content

Benchmark

GDM-MRCRv2: 8 needle

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

Models measured
68
Scores recorded
177
Sources
1
IQ weight
1.21

Score distribution

49.7%
Mean
50.1%
Median
20.2%
Std dev
24%
10th percentile
73.7%
90th percentile
12.3%
Min
12.3%94%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.21 for chat models, measured over 68 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
51Kimi K2.6enabled· Context Arena data64.6%
52DeepSeek V4 Pro 0813high· Context Arena data64.1%
53GPT-5.6 Lunamax· Context Arena data63.7%
54GLM 5.3 Flashhigh· Context Arena data63.4%
55Gemini 3.6 Flashlow· Context Arena data63%
56GPT 5.4medium· Context Arena data62.9%
57Muse Spark 1.2minimal· Context Arena data62.8%
58DeepSeek V4 Pro 0813max· Context Arena data62.7%
59GLM 5.3low· Context Arena data62.7%
60GLM 5.3 Flashmax· Context Arena data62.3%
61Doubao Seed 2.1 Turboenabled· Context Arena data62.1%
62GLM 5.1enabled· Context Arena data62%
63Claude Sonnet 5max· Context Arena data62%
64Muse Spark 1.1medium· Context Arena data60.6%
65Muse Spark 1.1minimal· Context Arena data60.1%
66GPT 5.4low· Context Arena data59.3%
67Kimi K2.5enabled· Context Arena data59.2%
68GLM 5.3 Flashlow· Context Arena data59.1%
69DeepSeek V4 Flash 0731max· Context Arena data58.9%
70Muse Spark 1.1low· Context Arena data58.9%
71Muse Spark 1.1high· Context Arena data58.5%
72Muse Spark 1.1xhigh· Context Arena data58.3%
73Qwen3.7 Flashenabled· Context Arena data58.1%
74GLM 5.2max· Context Arena data57.9%
75DeepSeek V4 Pro 0813low· Context Arena data57.3%
76DeepSeek V4 Flash 0731high· Context Arena data56.5%
77Qwen3.6 Plusenabled· Context Arena data56.1%
78DeepSeek V4 Proenabled· Context Arena data56%
79Gemini 3.1 Pro Previewhigh· Context Arena data53.8%
80Qwen3.6 27B· Context Arena data53.8%
81Kimi K2.5· Context Arena data53.3%
82Qwen3.6 35B A3B· Context Arena data53.2%
83Gemini 3.5 Flashhigh· Context Arena data53.2%
84Kimi K3max· Context Arena data53%
85Kimi K2.6· Context Arena data51.9%
86DeepSeek V4 Flashenabled· Context Arena data50.9%
87Nemotron 3 Super 120B A12Benabled· Context Arena data50.3%
88Ling 3.0 Flash· Context Arena data50.2%
89Qwen3.5 Plusenabled· Context Arena data50.1%
90GPT 5.4 Nanoxhigh· Context Arena data48.8%
91Gemini 3.1 Pro Previewlow· Context Arena data48.7%
92Claude Opus 4.6· Context Arena data48.2%
93Qwen3.6 Plus· Context Arena data47.9%
94Kimi K3high· Context Arena data47.1%
95Gemini 3 Flashhigh· Context Arena data46.8%
96Claude Sonnet 4.6· Context Arena data46.7%
97Qwen 3.8 27B· Context Arena data46.6%
98Gemini 3 Flashmedium· Context Arena data46.2%
99Gemini 3.5 Flashmedium· Context Arena data46.2%
100Qwen3.7 Flash· Context Arena data45.8%

Showing 51–100 of 177 results.