Skip to content

Benchmark

GDM-MRCRv2: 8 needle 16k

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
2.62

Score distribution

69.7%
Mean
79.3%
Median
25.6%
Std dev
34.2%
10th percentile
97.5%
90th percentile
13.6%
Min
13.6%100%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 2.62 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
101GLM 5.3 Flashlow· Context Arena data70.8%
102MiniMax M3enabled· Context Arena data69.3%
103Claude Opus 4.6· Context Arena data69.3%
104Qwen3.6 Plus· Context Arena data68.4%
105Qwen3.7 Flash· Context Arena data66%
106GPT 5.4 Nanoxhigh· Context Arena data64.6%
107Ling 3.0 Flash· Context Arena data63.7%
108Gemini 3.5 Flashlow· Context Arena data63.4%
109Kimi K2.5· Context Arena data62.5%
110GPT-5.4 minixhigh· Context Arena data62.5%
111Qwen3.6 Flash· Context Arena data62.4%
112MiMo V2.5 Proenabled· Context Arena data62.4%
113Gemini 3.1 Flash Litelow· Context Arena data62.4%
114Grok 4.20enabled· Context Arena data59.9%
115Kimi K2.6· Context Arena data58.7%
116Qwen3.7 Max· Context Arena data56.4%
117Qwen3.5 9Benabled· Context Arena data55.9%
118Inkling Smallmax· Context Arena data55.7%
119GPT-5.4 minilow· Context Arena data55.7%
120Qwen3.6 27B· Context Arena data55.1%
121GPT-5.4 minimedium· Context Arena data54.7%
122MiMo V2 Proenabled· Context Arena data54.7%
123Claude Sonnet 4.6· Context Arena data54.1%
124Qwen3.6 35B A3B· Context Arena data52.4%
125GPT 5.5· Context Arena data51.4%
126MiMo V2 Omnienabled· Context Arena data51.4%
127MiMo V2.5enabled· Context Arena data51.4%
128GPT 5.4 Nanohigh· Context Arena data48.9%
129Claude Haiku 4.5enabled· Context Arena data47.6%
130Doubao Seed 2.1 Turbo· Context Arena data46.7%
131Mistral Medium 3.5high· Context Arena data45.2%
132Qwen3.5 397B A17B· Context Arena data45%
133Nemotron 3 Super 120B A12Benabled· Context Arena data44.7%
134Inkling Smallmedium· Context Arena data44.5%
135Qwen 3.8 27B· Context Arena data44.1%
136Inkling Smallhigh· Context Arena data43%
137Inkling Smallminimal· Context Arena data41.7%
138Qwen3.5 35B A3B· Context Arena data41.4%
139Step 3.7 Flashmedium· Context Arena data40.9%
140Gemini 3.5 Flash· Context Arena data40.4%
141DeepSeek V4 Pro· Context Arena data40.2%
142Qwen3.5 Plus· Context Arena data40.1%
143Step 3.7 Flashlow· Context Arena data39.9%
144Step 3.7 Flashhigh· Context Arena data39.5%
145Inkling Smalllow· Context Arena data39.4%
146Qwen3.5 122B-A10B· Context Arena data39%
147Gemini 3.6 Flashminimal· Context Arena data38.1%
148GPT 5.4 Nanomedium· Context Arena data38%
149Qwen3.5 27B· Context Arena data37.9%
150Claude Opus 4.7· Context Arena data37.5%

Showing 101–150 of 177 results.