Fastest AI models (seconds per correct answer)

Ranked by wall-seconds per correct answer. For interactive and high-throughput workloads, latency per useful result — not raw tokens/sec — is what users feel. Models that think a lot to get there pay for it here.

Ranking 82 models across the meo-tested roster · as of 2026-10-19.

#	Model	Lab	sec/correct
1	Amazon: Nova Lite 1.0	amazon	5.31
2	Anthropic: Claude 3 Haiku	anthropic	5.77
3	Perplexity: Sonar	perplexity	6.93
4	OpenAI: GPT-6.1 Sol	openai	7.34
5	Anthropic: Claude Opus 5.5	anthropic	7.72
6	OpenAI: GPT-4o-mini	openai	11
7	OpenAI: GPT-4.1 Nano	openai	13
8	Anthropic: Claude Fable 5	anthropic	14
9	NVIDIA: Nemotron Nano 12B 2 VL (free)	nvidia	15
10	OpenAI: GPT-5.4 Nano	openai	16
11	Perceptron: Perceptron Mk1	perceptron	18
12	Mistral: Ministral 3 3B 2512	mistralai	18
13	Mistral: Mistral Large 3 2512	mistralai	18
14	SpaceXAI: Grok 4.3	x-ai	20
15	Anthropic: Claude Opus 4.8	anthropic	21
16	OpenAI: GPT-4.1 Mini	openai	22
17	Mistral: Ministral 3 8B 2512	mistralai	22
18	OpenAI: GPT-5.1-Codex-Mini	openai	22
19	Google: Gemma 3 12B	google	22
20	OpenAI: GPT-5.5	openai	22
21	Meta: Llama 4 Scout	meta-llama	23
22	Mistral: Ministral 3 14B 2512	mistralai	25
23	inclusionAI: Ring-2.6-1T	inclusionai	27
24	MiniMax: MiniMax-01	minimax	27
25	Google: Gemini 3.5 Flash	google	28
26	MoonshotAI: Kimi K3	moonshotai	32
27	Google: Gemma 3 27B	google	32
28	Google: Gemini 4 Argon	google	33
29	Qwen: Qwen2.5 VL 72B Instruct	qwen	35
30	Google: Gemini 3.1 Flash Lite Preview	google	35
31	Google: Gemma 3 4B	google	36
32	Nex AGI: Nex-N2-Pro (free)	nex-agi	37
33	Meta: Llama 4 Maverick	meta-llama	38
34	Mistral: Mistral Small 3.2 24B	mistralai	42
35	Google: Gemini 3.1 Pro Preview	google	44
36	Google: Gemini 2.5 Flash Lite	google	44
37	Qwen: Qwen3 VL 32B Instruct	qwen	45
38	Google: Gemini 3.1 Flash Lite	google	47
39	ByteDance Seed: Seed 1.6 Flash	bytedance-seed	53
40	Owl Alpha	openrouter	53
41	NVIDIA: Nemotron 3 Nano Omni (free)	nvidia	54
42	Mistral: Mistral Small 4	mistralai	55
43	OpenAI: GPT-5.5 Pro	openai	56
44	Xiaomi: MiMo-V2.5	xiaomi	57
45	OpenAI: GPT-5 Mini	openai	58
46	Qwen: Qwen3 VL 235B A22B Instruct	qwen	61
47	Qwen: Qwen3 VL 8B Instruct	qwen	63
48	Qwen: Qwen3 VL 30B A3B Instruct	qwen	64
49	Z.ai: GLM 4.6V	z-ai	64
50	NVIDIA: Nemotron 3 Ultra	nvidia	72
51	MiniMax: MiniMax M3	minimax	73
52	Meta: Llama 3.2 11B Vision Instruct	meta-llama	74
53	OpenAI: GPT-5 Nano	openai	80
54	MoonshotAI: Kimi K2.7 Code	moonshotai	88
55	Qwen: Qwen3.7 Max	qwen	102
56	Arcee AI: Trinity Large Thinking	arcee-ai	102
57	ByteDance Seed: Seed 1.6	bytedance-seed	104
58	Z.ai: GLM 5.2	z-ai	110
59	Qwen: Qwen3 VL 8B Thinking	qwen	111
60	Qwen: Qwen3.6 35B A3B	qwen	124
61	Qwen: Qwen3.7 Plus	qwen	128
62	Xiaomi: MiMo-V2.5-Pro	xiaomi	129
63	StepFun: Step 3.7 Flash	stepfun	131
64	DeepSeek: DeepSeek V4 Pro 0423	deepseek	131
65	Qwen: Qwen3.6 Flash	qwen	147
66	SpaceXAI: Grok 4.7	x-ai	151
67	MoonshotAI: Kimi K2.6	moonshotai	159
68	Qwen: Qwen3.5-Flash	qwen	165
69	DeepSeek: DeepSeek V4 Flash 0423	deepseek	169
70	Z.ai: GLM 5.1	z-ai	170
71	Baidu: ERNIE 4.5 VL 424B A47B	baidu	192
72	Tencent: Hy3 preview	tencent	216
73	Qwen: Qwen3.5 Plus 2026-02-15	qwen	225
74	Google: Gemma 4 26B A4B	google	258
75	Google: Gemma 4 31B	google	266
76	Qwen: Qwen3 VL 30B A3B Thinking	qwen	278
77	ByteDance Seed: Seed-2.0-Lite	bytedance-seed	285
78	Qwen: Qwen3.5-35B-A3B	qwen	290
79	Qwen: Qwen3.5-27B	qwen	293
80	Qwen: Qwen3.6 Plus	qwen	321
81	ByteDance Seed: Seed-2.0-Mini	bytedance-seed	382
82	Qwen: Qwen3.5-9B	qwen	476

← All rankings Methodology & 𝕍 →