Skip to content

Benchmark

GDM-MRCRv2: 8 needle 64k

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
1.74

Score distribution

50.4%
Mean
49.6%
Median
24.4%
Std dev
20%
10th percentile
84.1%
90th percentile
8.9%
Min
8.9%100%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.74 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
1Claude Opus 5max· Context Arena data100%
2GPT-5.6 Solmax· Context Arena data98.8%
3Qwen 3.8 27Bmedium· Context Arena data97.9%
4Qwen3.8 Maxxhigh· Context Arena data96.5%
5Qwen 3.8 27Blow· Context Arena data93.3%
6Qwen3.8 Maxlow· Context Arena data93.3%
7Qwen3.8 Maxmedium· Context Arena data93.3%
8Gemini 3.7 Flashhigh· Context Arena data93.1%
9GPT 5.5xhigh· Context Arena data90.9%
10Claude Opus 4.8max· Context Arena data90.7%
11GPT-5.6 Terramax· Context Arena data89.7%
12GLM 5.3high· Context Arena data89.6%
13Qwen3.6 Max Previewenabled· Context Arena data87.5%
14Qwen 3.8 27Bxhigh· Context Arena data87.4%
15Gemini 3.7 Flashmedium· Context Arena data86.3%
16GPT 5.5high· Context Arena data86.2%
17GPT 5.5low· Context Arena data85.2%
18GPT 5.5medium· Context Arena data85.1%
19DeepSeek V4 Pro 0813high· Context Arena data84.1%
20Claude Sonnet 4.6medium· Context Arena data81.8%
21GLM 5.3max· Context Arena data81.3%
22Gemini 3.7 Flashlow· Context Arena data80.6%
23GPT 5.4xhigh· Context Arena data80.6%
24Gemini 3.6 Flashhigh· Context Arena data80.6%
25Claude Opus 4.6medium· Context Arena data79.9%
26Claude Opus 4.6low· Context Arena data79.8%
27Claude Sonnet 4.6low· Context Arena data79.6%
28Claude Sonnet 4.6high· Context Arena data79.4%
29DeepSeek V4 Pro 0813max· Context Arena data77.3%
30Gemini 3.6 Flashmedium· Context Arena data77.1%
31Grok 4.6medium· Context Arena data77%
32GLM 5.3 Flashhigh· Context Arena data75.8%
33GLM 5.3low· Context Arena data75.6%
34GPT 5.4high· Context Arena data75%
35GLM 5.3 Flashlow· Context Arena data74.9%
36Gemini 3.1 Pro Previewhigh· Context Arena data74.9%
37Claude Opus 4.6high· Context Arena data74.3%
38Muse Spark 1.2high· Context Arena data72.6%
39DeepSeek V4 Flash 0731high· Context Arena data72.5%
40Qwen3.7 Flashenabled· Context Arena data72.4%
41Grok 4.6high· Context Arena data71.7%
42Grok 4.6xhigh· Context Arena data71.5%
43GPT-5.6 Lunamax· Context Arena data71.5%
44Qwen3.6 27Benabled· Context Arena data70.5%
45Qwen3.6 35B A3Benabled· Context Arena data70.4%
46GLM 5.3 Flashmax· Context Arena data70.1%
47Muse Spark 1.2xhigh· Context Arena data69%
48DeepSeek V4 Pro 0813low· Context Arena data68.1%
49Muse Spark 1.2minimal· Context Arena data68.1%
50Claude Sonnet 5max· Context Arena data68%

Showing 1–50 of 177 results.