Skip to content

Benchmark

RULER 4K

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

Models measured
27
Scores recorded
27
Sources
1
IQ weight
No weight published

Score distribution

94.8%
Mean
95.6%
Median
3%
Std dev
92.2%
10th percentile
96.9%
90th percentile
85.6%
Min
85.6%98.4%
The band covers the middle 80% of scores. The line marks the median.

Results

The source marks these results as descriptive, not capability scores, so we list them by name instead of ranking them.

ModelScore
C4AI Command R 08 2024Command-R-0824, 32B· RULER official results data94.7%
DBRX-instructDBRX, 36B/132B· RULER official results data95.1%
Gemini 1.5 ProGemini-1.5-pro· RULER official results data96.7%
GPT 4 1106 PreviewGPT-4-1106-preview· RULER official results data96.6%
Jamba 1.5 LargeJamba-1.5-large, 94B/398B· RULER official results data· Reported by the model's developer. The source relaying it did not verify it.96.7%
Jamba 1.5 MiniJamba-1.5-mini, 12B/52B· RULER official results data95.6%
Llama 3.1 70B InstructLlama3.1, 70B· RULER official results data96.5%
Llama 3.1 8B InstructLlama3.1, 8B· RULER official results data95.5%
Llama-2-7B-32K-InstructTogether, 7B· RULER official results data88.2%
Llama-2-7b-chat-hfLlama2, 7B· RULER official results data85.6%
Mistral 7B Instruct v0.2Mistral, 7B· RULER official results data93.6%
Mistral Large Instruct 2411Mistral-Large-2411, 123B· RULER official results data96.4%
Mistral NemoMistral-Nemo· RULER official results data87.8%
Mistral-Large-Instruct-2407Mistral-Large-2407, 123B· RULER official results data96.2%
Mixtral 8x22B Instruct v0.1Mixtral-8x22B, 39B/141B· RULER official results data95.6%
Mixtral 8x7B Instruct v0.1Mixtral-8x7B, 12.9B/46.7B· RULER official results data94.9%
Phi-3 Medium 128K InstructPhi3-medium, 14B· RULER official results data93.3%
Phi-3 Mini 128K InstructPhi3-mini, 3.8B· RULER official results data92.2%
Qwen 3 235B A22BQwen3-235B-A22B, 235B· RULER official results data· Reported by the model's developer. The source relaying it did not verify it.97.7%
Qwen1.5-72B-ChatQwen1.5, 72B· RULER official results data94.9%
Qwen2 72B InstructQwen2, 72B· RULER official results data96.9%
Qwen3 14BQwen3-14B, 14B· RULER official results data· Reported by the model's developer. The source relaying it did not verify it.98%
Qwen3 30B A3BQwen3-30B-A3B, 30B· RULER official results data· Reported by the model's developer. The source relaying it did not verify it.96.5%
Qwen3 32BQwen3-32B, 32B· RULER official results data· Reported by the model's developer. The source relaying it did not verify it.98.4%
Qwen3 4BQwen3-4B, 4B· RULER official results data· Reported by the model's developer. The source relaying it did not verify it.95.1%
Qwen3 8BQwen3-8B, 8B· RULER official results data· Reported by the model's developer. The source relaying it did not verify it.96.3%
Yi-34BYi, 34B, 200k· RULER official results data93.3%

Data credits: NVIDIA RULER, Hsieh et al. Official README numeric table; no separate results licence asserted.