Skip to content

Benchmark

GDM-MRCRv2: 8 needle 128k

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
1.80

Score distribution

40.7%
Mean
39.5%
Median
23%
Std dev
13.2%
10th percentile
72.6%
90th percentile
0.51%
Min
0.51%92.4%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.80 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
51GLM 5.3 Flashlow· Context Arena data55.4%
52GPT 5.4low· Context Arena data54.6%
53Claude Sonnet 5max· Context Arena data52.9%
54Grok 4.6xhigh· Context Arena data52.6%
55Qwen3.6 27B· Context Arena data52.5%
56Ring 2.6 1Thigh· Context Arena data52.1%
57Qwen3.7 Flash· Context Arena data51.9%
58Gemini 3.5 Flashmedium· Context Arena data51.8%
59Muse Spark 1.1high· Context Arena data51.8%
60GLM 5.2max· Context Arena data51.6%
61Muse Spark 1.2minimal· Context Arena data50.8%
62Muse Spark 1.1minimal· Context Arena data49.8%
63Ring 2.6 1Txhigh· Context Arena data49.8%
64Claude Sonnet 4.6· Context Arena data49%
65Grok 4.6high· Context Arena data49%
66Qwen3.6 Plus· Context Arena data48.9%
67Qwen3.6 35B A3B· Context Arena data48.9%
68Gemini 3.5 Flash Litemedium· Context Arena data48.8%
69DeepSeek V4 Proenabled· Context Arena data48.4%
70DeepSeek V4 Flash 0731low· Context Arena data47.4%
71Muse Spark 1.1medium· Context Arena data47.2%
72DeepSeek V4 Flash 0731high· Context Arena data47.1%
73Muse Spark 1.1low· Context Arena data47.1%
74Muse Spark 1.2xhigh· Context Arena data47%
75Grok 4.6medium· Context Arena data46.2%
76Ling 3.0 Flashenabled· Context Arena data45.4%
77Qwen3.6 Flash· Context Arena data45.2%
78Qwen3.5 Plusenabled· Context Arena data44.3%
79DeepSeek V4 Flashenabled· Context Arena data43.9%
80Gemini 3.1 Pro Previewlow· Context Arena data42.5%
81Grok 4.6low· Context Arena data42.3%
82Qwen3.5 27Benabled· Context Arena data42.3%
83Gemini 3.5 Flash Litelow· Context Arena data42.1%
84Qwen3.5 397B A17Benabled· Context Arena data42%
85Qwen3.7 Max· Context Arena data41.4%
86Muse Spark 1.1xhigh· Context Arena data41.2%
87DeepSeek V4 Pro 0813low· Context Arena data39.6%
88Kimi K3high· Context Arena data39.5%
89Kimi K2.6· Context Arena data39.5%
90Kimi K2.6enabled· Context Arena data39%
91Kimi K3max· Context Arena data38.3%
92Nemotron 3 Super 120B A12Benabled· Context Arena data38.1%
93Claude Opus 4.6· Context Arena data37.5%
94Ling 3.0 Flash· Context Arena data36.1%
95GPT-5.4 minimedium· Context Arena data34.8%
96Gemini 3 Flashhigh· Context Arena data34.8%
97Qwen3.5 122B-A10Benabled· Context Arena data34.7%
98Qwen 3.8 27B· Context Arena data34.7%
99Kimi K2.5enabled· Context Arena data33%
100GPT-5.4 minixhigh· Context Arena data33%

Showing 51–100 of 177 results.