IFBench

Precise instruction following on unseen constraints. Run by Artificial Analysis.

Published by

Artificial Analysis

Category

Instruction following

Index weight

1.0

Models

401

Data as of

9 Sept 2026

Source: Artificial Analysis (artificialanalysis.ai). Data taken from the public leaderboard.

1. 1Grok 4.383.3%
2. 2Grok 4.2082.9%
3. 3MiniMax-M382.9%
4. 4Nemotron 3 Ultra 550B A55B81.4%
5. 5Grok 4.381.3%
6. 6Grok 4.381.0%
7. 7Qwen3 780.5%
8. 8Nemotron Cascade 2 30B A3B80.4%
9. 9MiMo-V2.5-Pro79.9%
10. 10Nova 2.0 Pro Preview79.6%
11. 11DeepSeek V4 Flash79.2%
12. 12Nova 2.0 Pro Preview79.0%
13. 13Qwen3.5 397B-A17B78.8%
14. 14Gemini 3 Flash78.0%
15. 15Qwen3.7 Plus78.0%

SearchAll providersAI21 LabsAlibabaAllen Institute for AIAllenaiAmazonAnt GroupAnt-groupAnthropicArcee AIBaiduByteDanceChina MobileCohereDeep CogitoDeepSeekGoogleIBMInceptionKimiKorea TelecomKwaiKATLG AI ResearchLiquid AILongCatMBZUAI Institute of Foundation ModelsMetaMicrosoftMiniMaxMistral AIMoonshot AIMotif TechnologiesMultiverse ComputingNanbeigeNaverNex AGINous ResearchNVIDIAOpenAIOpenBMBPrime IntellectQwenReka AISarvamServiceNowStepFunSwiss AI InitiativeTII UAETrillion LabsUnknownUpstagexAIXiaomiZhipu AIOpen weights401 of 401

#	Model	Score ↓	Run ↕	Index ↕	Published ↕
1	Grok 4.3xAI	83.3%	–	60.7	–
2	Grok 4.20xAI	82.9%	0309	57.9	–
3	MiniMax-M3MiniMaxopen	82.9%	MiniMax-M3	56.9	–
4	Nemotron 3 Ultra 550B A55BNVIDIAopen	81.4%	Nemotron 3 Ultra	58.3	–
5	Grok 4.3xAI	81.3%	Grok 4.3 (high)	57.8	–
6	Grok 4.3xAI	81.0%	–	59.5	–
7	Qwen3 7Alibaba	80.5%	–	58.9	–
8	Nemotron Cascade 2 30B A3BNVIDIAopen	80.4%	Nemotron Cascade 2 30B A3B	51.0	–
9	XMiMo-V2.5-ProXiaomiopen	79.9%	–	59.0	–
10	Nova 2.0 Pro PreviewAmazon	79.6%	–	51.3	–
11	DeepSeek V4 FlashDeepSeekopen	79.2%	0420	59.1	–
12	Nova 2.0 Pro PreviewAmazon	79.0%	–	52.3	–
13	Qwen3.5 397B-A17BAlibabaopen	78.8%	Qwen3.5 397B A17B	58.6	–
14	Gemini 3 FlashGoogle	78.0%	–	61.5	–
15	Qwen3.7 PlusAlibaba	78.0%	Qwen3.7 Plus	59.7	–
16	GPT-5.2 CodexOpenAI	77.6%	GPT-5.2 Codex (xhigh)	60.7	–
17	Gemini 3.1 Flash LiteGoogle	77.2%	–	54.3	–
18	Gemini 3.1 ProGoogle	77.1%	–	63.9	–
19	Qwen3 6Alibaba	76.6%	–	62.0	–
20	DeepSeek V4 ProDeepSeekopen	76.5%	0424	59.7	–
21	Gemini 3.5 FlashGoogle	76.3%	Gemini 3.5 Flash	60.7	–
22	GLM-5.1Zhipu AIopen	76.3%	GLM-5.1	60.1	–
23	Kimi K2.6Moonshot AIopen	76.0%	Kimi K2.6	60.6	–
24	Muse SparkMeta	75.9%	–	65.4	–
25	GPT-5.4 nanoOpenAI	75.9%	GPT-5.4 nano (xhigh)	53.9	–
26	GPT-5.5OpenAI	75.8%	GPT-5.5 (xhigh)	66.6	–
27	MiniMax-M2.7MiniMaxopen	75.7%	MiniMax-M2.7	56.5	–
28	Qwen3.5 122B-A10BAlibabaopen	75.7%	Qwen3.5 122B A10B	56.0	–
29	Gemma-4-31B-ITNVIDIAopen	75.6%	Gemma 4 31B	55.4	–
30	Qwen3.5 27BAlibabaopen	75.6%	Qwen3.5 27B	55.4	–
31	GPT-5.2OpenAI	75.4%	GPT-5.2 (xhigh)	61.5	–
32	GPT-5 miniOpenAI	75.4%	GPT-5 mini (high)	55.8	–
33	GPT-5.3 CodexOpenAI	75.4%	GPT-5.3 Codex (xhigh)	62.4	–
34	Qwen3.6 PlusAlibabaopen	75.2%	Qwen3.6 Plus	59.0	–
35	Gemini 3.5 FlashGoogle	74.6%	–	63.9	–
36	GPT-5-CodexOpenAI	74.2%	GPT-5 Codex (high)	58.1	–
37	GPT-5.4OpenAI	74.0%	GPT-5.4 (xhigh)	63.4	–
38	Command A+Cohereopen	74.0%	–	53.9	–
39	Gemma 4 12BGoogleopen	73.5%	Gemma 4 12B	52.1	–
40	DeepSeek V4 FlashDeepSeekopen	73.5%	high, 0420	59.3	–
41	GLM-5.2Zhipu AIopen	73.3%	GLM-5.2 (max)	57.4	–
42	GPT-5.4 miniOpenAI	73.3%	GPT-5.4 mini (xhigh)	54.8	–
43	GLM-5-TurboZhipu AI	73.2%	–	58.1	–
44	GPT-5OpenAI	73.1%	GPT-5 (high)	59.5	–
45	GPT-5.1OpenAI	72.9%	GPT-5.1 (high)	59.4	–
46	GPT-5.6 SolOpenAI	72.7%	GPT-5.6 Sol (max)	65.3	–
47	Qwen3.5 35B-A3BAlibabaopen	72.5%	Qwen3.5 35B A3B	53.4	–
48	Gemma 4 26B A4BGoogleopen	72.5%	Gemma 4 26B A4B	54.0	–
49	GLM-5Zhipu AIopen	72.3%	GLM-5	59.0	–
50	MiniMax-M2MiniMaxopen	72.3%	MiniMax-M2	55.1	–
51	XMiMo-V2-FlashXiaomiopen	71.8%	feb-2026	53.3	–
52	GPT-5.5OpenAI	71.6%	–	66.3	–
53	MiniMax-M2.5MiniMaxopen	71.6%	MiniMax-M2.5	53.5	–
54	GPT-5.5 InstantOpenAI	71.5%	–	60.4	–
55	Nemotron 3 SuperNVIDIAopen	71.5%	Nemotron 3 Super	53.6	–
56	o3OpenAI	71.4%	o3	60.3	–
57	DeepSeek V4 ProDeepSeekopen	71.3%	high, 0424	60.0	–
58	GPT-5.6 TerraOpenAI	71.2%	GPT-5.6 Terra (max)	62.3	–
59	Solar Pro 3Upstage	71.2%	Solar Pro 3	44.6	–
60	GPT-5 miniOpenAI	71.2%	–	54.5	–

Show 100 more