Best LLMs for Instruction Following
The best LLM for instruction following is MiniMax M3, satisfying 82.9% of IFBench's verifiable prompt constraints. Grok 4.3 (81.3%) and Grok 4.20 (81.2%) round out the top three.
- 1
minimax-m3ReasoningToolsJSON+144.4 intel · $0.300/M · 101 t/s82.9%IFBench - 2
grok-4.3ReasoningToolsJSON+137.6 intel · $1.25/M · 1M ctx81.3%IFBench - 3grok-4.20ReasoningToolsJSON+137.0 intel · $1.25/M · 2M ctx81.2%IFBench
- 4
qwen3.7-maxReasoningToolsJSON46.0 intel · $1.48/M · 207 t/s80.5%IFBench - 5
mimo-v2.5-proReasoningToolsJSON42.2 intel · $0.435/M · 65 t/s79.9%IFBench - 6qwen3.5-397b-a17bReasoningToolsJSON+133.7 intel · $0.390/M · 69 t/s78.8%IFBench
- 7qwen3.7-plusReasoningToolsJSON+139.0 intel · $0.320/M · 12 t/s78.0%IFBench
- 8
gpt-5.2-codexReasoningToolsJSON+140.1 intel · $1.75/M · 400K ctx77.6%IFBench - 9
gemini-3.1-flash-liteReasoningToolsJSON+225.0 intel · $0.250/M · 337 t/s77.2%IFBench - 10gemini-3.1-flash-lite-previewReasoningToolsJSON+225.0 intel · $0.250/M · 337 t/s77.2%IFBench
- 11gemini-3.1-pro-previewReasoningToolsJSON+246.5 intel · $2.00/M · 136 t/s77.1%IFBench
- 12qwen3.6-max-previewReasoningToolsJSON40.0 intel · $1.03/M · 262K ctx76.6%IFBench
- 13
deepseek-v4-proReasoningToolsJSON44.3 intel · $0.435/M · 67 t/s76.5%IFBench - 14gemini-3.5-flashReasoningToolsJSON+250.2 intel · $1.50/M · 275 t/s76.3%IFBench
- 15
glm-5.1ReasoningToolsJSON40.2 intel · $0.966/M · 205K ctx76.3%IFBench - 16
kimi-k2.6ReasoningToolsJSON+144.2 intel · $0.589/M · 262K ctx76.0%IFBench - 17gpt-5.4-nanoReasoningToolsJSON+138.2 intel · $0.200/M · 400K ctx75.9%IFBench
- 18gpt-5.5ReasoningToolsJSON+154.8 intel · $5.00/M · 1.1M ctx75.9%IFBench
- 19minimax-m2.7ReasoningToolsJSON38.1 intel · $0.270/M · 205K ctx75.7%IFBench
- 20qwen3.5-122b-a10bReasoningToolsJSON+132.3 intel · $0.260/M · 127 t/s75.7%IFBench
- 21qwen3.5-27bReasoningToolsJSON+133.8 intel · $0.195/M · 262K ctx75.6%IFBench
- 22gpt-5.2ReasoningToolsJSON+142.2 intel · $1.75/M · 74 t/s75.4%IFBench
- 23gpt-5-miniReasoningToolsJSON+125.3 intel · $0.250/M · 84 t/s75.4%IFBench
- 24gpt-5.3-codexReasoningToolsJSON+144.3 intel · $1.75/M · 122 t/s75.4%IFBench
- 25qwen3.6-plusReasoningToolsJSON+139.6 intel · $0.325/M · 55 t/s75.2%IFBench
How this is ranked
AI models ranked by IFBench — how reliably a model obeys explicit, verifiable constraints in the prompt (format, length, inclusion and exclusion rules). High scores mean fewer retries and less prompt-wrangling in production, which often matters more than raw intelligence for pipelines.
Frequently asked
Which LLM follows instructions most reliably?
The best LLM for instruction following is MiniMax M3, satisfying 82.9% of IFBench's verifiable prompt constraints. Grok 4.3 (81.3%) and Grok 4.20 (81.2%) round out the top three.
Which AI model follows instructions best?
The best AI model for instruction following is MiniMax M3, satisfying 82.9% of IFBench's verifiable prompt constraints. Grok 4.3 (81.3%) and Grok 4.20 (81.2%) round out the top three.
What's a good alternative to MiniMax M3?
Grok 4.3 (81.3%) is the closest alternative on this metric, followed by Grok 4.20 (81.2%). See the full ranking above for the tradeoffs.