Instruction-Following Eval (IFEval)
We show this table for reference; we do not rank on it.
LFM2.5-24B-A2B has the highest reported IFEval score at 96.8%, ahead of LFM3-1.7B (96.7%) and Flint (95.1%) among 70 models. We compile the table from provider self-reports and secondary reports and keep it for display only; it does not affect overall rankings.
A benchmark of 541 prompts built from 25 verifiable instruction types. It tests whether a model follows checkable constraints such as keyword, length, casing, and response-format requirements.
Benchmark score on IFEval — October 29, 2026
We compile the IFEval rows from provider self-reports and secondary reports. LFM2.5-24B-A2B leads the table at 96.8%, followed by LFM3-1.7B (96.7%) and Flint (95.1%). We do not use these results to rank models overall.
LFM2.5-24B-A2B
LiquidAI
IFEval metric view not recorded
LFM3-1.7B
LiquidAI
IFEval metric view not recorded
Flint
Recursive
IFEval metric view not recorded
70 modelsInstruction FollowingStaleDisplay onlyUpdated October 29, 2026
Benchmark score table (70 models)
| Rank | Model / configuration | Score | Parameters (B) | Open / closed |
|---|---|---|---|---|
| 1 | LFM2.5-24B-A2BLiquidAIIFEval metric view not recorded | 96.8% | Not reported | Open |
| 2 | LFM3-1.7BLiquidAIIFEval metric view not recorded | 96.7% | Not reported | Open |
| 3 | FlintRecursiveIFEval metric view not recorded | 95.1% | Not reported | Open |
| 4 | Qwen3.5-27BAlibabaIFEval metric view not recorded | 95% | Not reported | Open |
| 5 | Agents-A1InternScienceIFEval metric view not recorded | 94.8% | Not reported | Open |
| 6 | Agents-A1-4BInternScienceIFEval metric view not recorded | 94.8% | Not reported | Open |
| 7 | Qwen3.7 PlusAlibabaIFEval metric view not recorded | 94.6% | Not reported | Closed |
| 8 | Qwen3.7 MaxAlibabaIFEval metric view not recorded | 94.3% | Not reported | Closed |
| 9 | Qwen3.6 PlusAlibabaIFEval metric view not recorded | 94.3% | Not reported | Closed |
| 10 | Kimi K2.5Moonshot AIIFEval metric view not recorded | 93.9% | Not reported | Open |
| 11 | dots3-note PreviewDots StudioIFEval metric view not recorded | 93.9% | Not reported | Open |
| 12 | o3-miniOpenAIIFEval metric view not recorded | 93.9% | Not reported | Closed |
| 13 | Spark-X2.6-4BSparkLLMIFEval metric view not recorded | 93.6% | Not reported | Open |
| 14 | Qwen3.5-122B-A10BAlibabaIFEval metric view not recorded | 93.4% | Not reported | Open |
| 15 | Claude 3.7 Sonnet ThinkingAnthropicIFEval metric view not recorded | 93.2% | Not reported | Closed |
| 16 | GLM-5Z.AIIFEval metric view not recorded | 92.6% | Not reported | Open |
| 17 | Qwen3.5 397BAlibabaIFEval metric view not recorded | 92.6% | Not reported | Open |
| 18 | K-EXAONE 2.0LG AI ResearchIFEval metric view not recorded | 92.4% | Not reported | Open |
| 19 | o1OpenAIIFEval metric view not recorded | 92.2% | Not reported | Closed |
| 20 | Llama 3.3 70BMetaIFEval metric view not recorded | 92.1% | Not reported | Open |
| 21 | Qwen3.5-35B-A3BAlibabaIFEval metric view not recorded | 91.9% | Not reported | Open |
| 22 | LFM2.5-8B-A1BLiquidAIIFEval metric view not recorded | 91.8% | Not reported | Open |
| 23 | Ternary Bonsai 2 27BPrism MLIFEval metric view not recorded | 91.3% | Not reported | Open |
| 24 | Claude Opus 4.5AnthropicIFEval metric view not recorded | 90.9% | Not reported | Closed |
| 25 | Claude 3.7 SonnetAnthropicIFEval metric view not recorded | 90.8% | Not reported | Closed |
| 26 | Mellum2.1-12B-A2.5B-ThinkingJetBrainsIFEval metric view not recorded | 90.6% | Not reported | Open |
| 27 | Kimi K2Moonshot AIIFEval metric view not recorded | 89.8% | Not reported | Closed |
| 28 | Nemotron Ultra 253BNVIDIAIFEval metric view not recorded | 89.5% | Not reported | Open |
| 29 | Qwen3 235B 2507AlibabaIFEval metric view not recorded | 88.7% | Not reported | Open |
| 30 | Llama 3.1 405BMetaIFEval metric view not recorded | 88.6% | Not reported | Open |
| 31 | GPT-4.1 miniOpenAIIFEval metric view not recorded | 88.5% | Not reported | Closed |
| 32 | Qwen3 235B 2507 (Reasoning)AlibabaIFEval metric view not recorded | 87.8% | Not reported | Open |
| 33 | Llama 3.1 70BMetaIFEval metric view not recorded | 87.5% | Not reported | Open |
| 34 | GPT-4.1OpenAIIFEval metric view not recorded | 87.4% | Not reported | Closed |
| 35 | MiniCPM5-2BOpenBMBIFEval metric view not recorded | 86.7% | Not reported | Open |
| 36 | DeepSeek V3DeepSeekIFEval metric view not recorded | 86.1% | Not reported | Open |
| 37 | Claude 3.5 HaikuAnthropicIFEval metric view not recorded | 85.9% | Not reported | Closed |
| 38 | ZAYA1-8BZyphraIFEval metric view not recorded | 85.6% | Not reported | Open |
| 39 | Qwen3 32B (Reasoning)AlibabaIFEval metric view not recorded | 85% | Not reported | Open |
| 40 | Phi-4-reasoning-plusMicrosoftIFEval metric view not recorded | 84.9% | Not reported | Open |
| 41 | Sarvam 105BSarvamIFEval metric view not recorded | 84.8% | Not reported | Open |
| 42 | QwQ-32BAlibabaIFEval metric view not recorded | 83.9% | Not reported | Open |
| 43 | Qwen3 235B-A22B (Reasoning)AlibabaIFEval metric view not recorded | 83.4% | Not reported | Open |
| 44 | Phi-4-reasoningMicrosoftIFEval metric view not recorded | 83.4% | Not reported | Open |
| 45 | DeepSeek-R1DeepSeekIFEval metric view not recorded | 83.3% | Not reported | Open |
| 46 | Qwen3 235B-A22BAlibabaIFEval metric view not recorded | 83.2% | Not reported | Open |
| 47 | Qwen3 32BAlibabaIFEval metric view not recorded | 83.2% | Not reported | Open |
| 48 | GPT-4.1 nanoOpenAIIFEval metric view not recorded | 83.2% | Not reported | Closed |
| 49 | OrcaCyber-Nano-1.0OrcaRouterIFEval metric view not recorded | 82.8% | Not reported | Open |
| 50 | LFM2.5-VL-3BLiquidAIIFEval metric view not recorded | 82.3% | Not reported | Open |
| 51 | Ternary Bonsai 8BPrism MLIFEval metric view not recorded | 81.8% | Not reported | Open |
| 52 | Kanana-2 3B InstructKakaoIFEval metric view not recorded | 81.0% | Not reported | Open |
| 53 | Celeris-1CelerisIFEval metric view not recorded | 80.8% | Not reported | Closed |
| 54 | MiniCPM5-1BOpenBMBIFEval metric view not recorded | 80.4% | Not reported | Open |
| 55 | Llama 3.1 8BMetaIFEval metric view not recorded | 80.4% | Not reported | Open |
| 56 | 1-bit Bonsai 8BPrism MLIFEval metric view not recorded | 79.8% | Not reported | Open |
| 57 | Granite-4.0-H-1BIBMIFEval metric view not recorded | 78.5% | Not reported | Open |
| 58 | Kanana-2 1.3B InstructKakaoIFEval metric view not recorded | 77.6% | Not reported | Open |
| 59 | Granite-4.0-1BIBMIFEval metric view not recorded | 77.4% | Not reported | Open |
| 60 | LFM2.5-350MLiquidAIIFEval metric view not recorded | 77.0% | Not reported | Open |
| 61 | Mellum2-12B-A2.5B-ThinkingJetBrainsIFEval metric view not recorded | 76.5% | Not reported | Open |
| 62 | Mellum2-12B-A2.5B-InstructJetBrainsIFEval metric view not recorded | 75.8% | Not reported | Open |
| 63 | Ternary Bonsai 4BPrism MLIFEval metric view not recorded | 72.1% | Not reported | Open |
| 64 | LFM2.5-230MLiquidAIIFEval metric view not recorded | 71.7% | Not reported | Open |
| 65 | Ternary Bonsai 1.7BPrism MLIFEval metric view not recorded | 70.1% | Not reported | Open |
| 66 | 1-bit Bonsai 4BPrism MLIFEval metric view not recorded | 69.6% | Not reported | Open |
| 67 | 1-bit Bonsai 1.7BPrism MLIFEval metric view not recorded | 63% | Not reported | Open |
| 68 | Granite-4.0-H-350MIBMIFEval metric view not recorded | 61.6% | Not reported | Open |
| 69 | LFM2.5-VL-450MLiquidAIIFEval metric view not recorded | 61.2% | Not reported | Open |
| 70 | Granite-4.0-350MIBMIFEval metric view not recorded | 55.4% | Not reported | Open |
How to read this leaderboard
Editorial review by Glevd · 2026-07-15
Use IFEval to judge compliance with objectively checkable constraints after matching both scoring choices: prompt versus instruction level, and strict versus loose. An unlabeled IFEval number is incomplete because the four metrics answer different questions and can produce different scores.
Operator receipt: 70 sourced rows are currently displayable on this page; the leading published row is LFM2.5-24B-A2B at 96.8%.
Honest limit: IFEval covers constraints that can be verified automatically. It does not establish factual accuracy, usefulness, judgment, tone, or multi-turn reliability, and strict checks can reject a substantively compliant answer because of formatting. The current table marks public rows whose prompt/instruction and strict/loose metric view is not recorded; do not treat those values as a precise normalized race.
Among the reported IFEval rows, LFM2.5-24B-A2B is first at 96.8%. The third row is 1.7 points behind. The broader top-10 range is 2.8 points, so many of the published results sit in a relatively narrow band.
70 models have been evaluated on IFEval. The benchmark falls in the Instruction Following category. IFEval is currently displayed for reference but excluded from the scoring formula, so it does not directly affect overall rankings.
About IFEval
Year
2023
Tasks
541 prompts across 25 instruction types
Format
Constrained generation
Difficulty
Instruction precision
IFEval applies deterministic checkers and reports four views of the same responses: prompt-level strict, instruction-level strict, prompt-level loose, and instruction-level loose. Prompt-level scoring requires every instruction in a prompt to pass; loose scoring applies limited normalization before checking.
Freshness and provenance
Version
IFEval 2023
Refresh cadence
Static
Staleness state
Stale
Question availability
Public benchmark set
BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.
Questions
What does IFEval measure?
IFEval measures whether a model follows checkable constraints in 541 prompts built from 25 instruction types. Examples include using required keywords, avoiding specified terms, meeting length limits, or producing a requested structure. Automatic checkers score each response, which makes the result reproducible for the constraints they can verify.
Which IFEval score should I compare?
Match all four metric choices before comparing models: prompt-level or instruction-level, and strict or loose. Prompt-level scoring requires every instruction in a prompt to pass. Loose scoring applies limited normalization to reduce formatting-related failures. An unlabeled IFEval number is not enough evidence for a precise comparison.
Does a high IFEval score mean a model is a good assistant?
No. A high IFEval score means the model obeyed more objectively checkable constraints under the selected metric. It does not show that the answer was factual, useful, well judged, stylistically appropriate, or reliable across a long conversation. Pair it with broader instruction tests and a workflow-specific prompt set.
Compare top models on IFEval
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 6,100+ subscribers.
One email each week. Unsubscribe anytime.