Skip to content

Benchmark

IFBench

Instruction following benchmarks measure how closely models follow instructions. Scores are percentages. Higher is better.

Models measured
51
Scores recorded
70
Sources
2
IQ weight
1.40

Score distribution

71.6%
Mean
74.2%
Median
11.2%
Std dev
59.2%
10th percentile
81.3%
90th percentile
21%
Min
21%82.8%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.40 for chat models, measured over 51 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
51Qwen3.5 35B A3B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim70.2%
52MiniMax M2.1· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim70%
53MiniMax M2.5official appendix· Reported by MiniMax· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.70%
54GPT OSS 120Bhigh· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim69.5%
55Mai Thinking 1· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim69%
56Mistral Medium 3.5· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim69%
57Step 3.5 FlashStep 3.5 technical report · Table 5 · vanilla· Reported by StepFun· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.67.4%
58Qwen3.5 9B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim64.5%
59Claude Fable 5Qwen3.8 官方模型卡对照表· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.63.5%
60Claude Opus 4.6Qwen3.8 Omni Flash 官方发布表 · max effort· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.62.5%
61Claude Opus 4.8Qwen3.8 官方模型卡对照表· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.62.2%
62Qwen3.5 4B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim59.2%
63LFM2.5 2.6B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim59.2%
64Step 3.5 FlashStep 3.5 technical report · Table 5 · PaCoRe· Reported by StepFun· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.56.8%
65Qwen3.5 Omni PlusQwen3.5-Omni release · no-thinking setting· Reported by Qwen, Qwen Team· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.52.6%
66Trinity Large ThinkingTrinity-Large-Thinking Model Card· Reported by Arcee AI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.52.3%
67Mistral Smalllatest· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim48%
68Qwen3.5 2B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim41.3%
69Qwen3.5 Omni FlashQwen3.5-Omni release · no-thinking setting· Reported by Qwen, Qwen Team· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.38.4%
70Qwen3.5 0.8B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim21%

Showing 51–70 of 70 results.

Data credits: Data by LLM Stats · BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links.