Which AI model is the most accurate?

Claude Opus 5.5 by Anthropic scored highest in our 2026-10-27 run, answering 68 of 80 questions correctly (85.0%), out of 78 models tested.

Which AI model is the fastest?

Nemotron 3.5 Super VL by NVIDIA had the highest median output speed in our 2026-10-27 run: 715 tokens per second.

Which AI model is the cheapest to run?

Granite 4.0 Micro by IBM had the lowest billed cost per question in our 2026-10-27 run: $0.000051, reasoning tokens included.

Which AI model responds the quickest?

Ministral 3 3B 2512 by Mistral had the shortest median time to first token in our 2026-10-27 run: 0.3 s.

Which AI model gives a complete answer the quickest?

Nemotron 3.5 Super VL by NVIDIA had the shortest median end-to-end time in our 2026-10-27 run: 1.5 seconds to stream a full answer of about 300 words.

Which AI model has the largest context window?

Llama 4 Scout by Meta accepts the most context of the models we test: 1,310,720 tokens, according to OpenRouter's catalogue.

What is the best open-weights AI model?

Muse Spark Open by Meta has the highest verification score among open-weights models we test: 77.5%.

How many AI models do you compare?

78 models from 28 labs, each tested through OpenRouter with the same questions and speed runs.

Models · tested 27 Oct 2026

Compare AI models: accuracy, speed, latency and price

78 language models from 28 labs, all called through the same public API with the same questions and speed runs. Select a model for its full results, compare any two, or see the best open-weights models.

Most accurateVerification score1Claude Opus 5.585.0%
2Gemini 4 Argon84.8%
3GPT-5.3-Codex82.5%
Fastest outputOutput speed1Nemotron 3.5 Super VL715tok/s
2Gemini 3.5 Flash Lite409tok/s
3Trinity Large Thinking329tok/s
Lowest latencyTime to first token1Ministral 3 3B 25120.3 s
2Ministral 3 8B 25120.3 s
3Hermes 4 405B0.3 s
Cheapest per questionCost per question1Granite 4.0 Micro$0.000051
2gpt-oss-20b$0.000066
3Granite 4.2 8B$0.000077
Quickest full responseEnd-to-end response time1Nemotron 3.5 Super VL1.5 s
2Ministral 3 3B 25121.6 s
3Nova Micro 1.02.1 s
Largest context windowContext window1Llama 4 Scout1.3Mtokens
2GPT-6 Astra1.1Mtokens
2GPT-5.6 Terra1.1Mtokens

Accuracy

Share of 80 questions answered correctly. Every model gets the same questions.

Verification score

Share of 80 questions answered correctly · higher is better

52 more models · see the full leaderboard

Score vs. cost per question

Filled dots: no cheaper model scores as high

Results by category

Questions answered correctly in each category

Model	Score	Science30 questions	Engineering15 questions	Medicine11 questions	Economics9 questions	Law6 questions	Humanities & social science9 questions
Claude Opus 5.5Anthropic	85.0	28/30	12/15	7/11	8/9	5/6	8/9
Gemini 4 ArgonGoogle	84.8	25/30	13/14	7/11	8/9	6/6	8/9
GPT-5.3-CodexOpenAI	82.5	28/30	11/15	7/11	8/9	4/6	8/9
Gemini 3.8 FlashGoogle	82.5	26/30	12/15	6/11	8/9	6/6	8/9
Claude Fable 5.1Anthropic	81.2	25/30	13/15	7/11	7/9	5/6	8/9
Grok 4.7SpaceXAI	80.0	27/30	11/15	6/11	8/9	4/6	8/9
Muse WonderMeta	78.8	26/30	11/15	5/11	8/9	4/6	9/9
Claude Sonnet 5.5Anthropic	78.8	24/30	13/15	7/11	8/9	3/6	8/9
Gemini 3.1 Pro PreviewGoogle	77.5	26/30	10/15	7/11	8/9	3/6	8/9
Muse Spark OpenMeta	77.5	25/30	11/15	6/11	8/9	4/6	8/9
Kimi K3.1 ProMoonshotAI	77.5	26/30	11/15	6/11	8/9	3/6	8/9
GPT-6 AstraOpenAI	76.2	26/30	10/15	6/11	8/9	3/6	8/9
GPT-6.1 Terra ProOpenAI	76.2	26/30	10/15	6/11	8/9	3/6	8/9
GPT-6.1 SolOpenAI	76.2	27/30	10/15	5/11	8/9	3/6	8/9
GPT-5.6 TerraOpenAI	75.0	26/30	11/15	4/11	7/9	4/6	8/9
Gemini 3.5 Flash LiteGoogle	75.0	23/30	10/15	7/11	7/9	4/6	9/9
Qwen3.8 Max (0902)Qwen	75.0	24/30	9/15	6/11	8/9	4/6	9/9
GPT-6.1 TerraOpenAI	73.8	26/30	9/15	6/11	7/9	3/6	8/9
DeepSeek V4.1 Flash 1017DeepSeek	73.8	26/30	9/15	4/11	7/9	4/6	9/9
GPT-6 SolOpenAI	73.8	27/30	8/15	5/11	7/9	4/6	8/9

Cost

What it actually cost to run each model, billed by OpenRouter, next to the list price. Reasoning tokens are billed as output, so models that think longer cost more per question.

Cost per question

Average billed USD per question, reasoning included · lower is better

1Granite 4.0 Micro$0.000051
2gpt-oss-20b$0.000066
3Granite 4.2 8B$0.000077
4Nova Micro 1.0$0.000087
5Mercury 2.5$0.00011
6Phi 4$0.00012
7gpt-oss-120b$0.00013
8GLM 5.3 Flash$0.00015
9Ministral 3 3B 2512$0.00019
10GPT-6 Luna$0.00025
11Ministral 3 8B 2512$0.00027
12Ministral 3 14B 2512$0.00031
13Llama 4 Scout$0.00037
14Inkling Small$0.00047
15Claude Haiku 5.5$0.00048
16Nemotron 3.5 Lightning$0.00049
17Solar Mini 4$0.00053
18Llama 3.3 70B Instruct$0.00053
19MiMo-V2.6-Flash$0.00056
20Llama 4 Maverick$0.00059

52 more models · see the full leaderboard

List price per 1M output tokens

OpenRouter list price · lower is cheaper

1gpt-oss-20b$0.018 in$0.09
2Ministral 3 3B 2512$0.1 in$0.1
3Granite 4.0 Micro$0.017 in$0.112
4Nova Micro 1.0$0.035 in$0.14
5Phi 4$0.07 in$0.14
6Mercury 2.5$0.04 in$0.15
7Ministral 3 8B 2512$0.15 in$0.15
8Ling 3.0 Flash VL$0.06 in$0.18
9Reka Flash 3$0.1 in$0.2
10Solar Mini 4$0.05 in$0.2
11Nemotron 3.5 Lightning$0.07 in$0.2
12Ministral 3 14B 2512$0.2 in$0.2
13Ling 3.1 Flash$0.075 in$0.225
14Granite 4.2 8B$0.06 in$0.25
15MiMo-V2.6-Flash$0.14 in$0.28
16Llama 4 Scout$0.1 in$0.3
17Gemma 4 26B A4B$0.09 in$0.3
18Llama 3.3 70B Instruct$0.1 in$0.32
19Gemma 4 31B$0.2 in$0.4
20Nemotron 3.5 Super VL$0.15 in$0.5

Speed

Output tokens per second after the first token, median of 3 streamed runs of the same prompt.

Output speed

Median output tokens per second · higher is better

52 more models · see the full leaderboard

Output speed vs. cost per question

Fast and cheap sits top left

Latency

Seconds from sending a request to receiving the first token. Reasoning models think before answering, and that time counts.

Time to first token

Median seconds to the first token · lower is better

54 more models · see the full leaderboard

Latency vs. output speed

Quick to start and fast to finish sits top left

End-to-end response time

Seconds from sending the request to receiving the last token of a full answer of about 300 words: waiting, thinking and writing combined.

End-to-end response time

Median seconds for the full answer · lower is better

54 more models · see the full leaderboard

Score vs. end-to-end response time

Accurate and quick sits top left

Context window

The most tokens of input and output a model accepts in one request, from OpenRouter's catalogue. A larger window fits longer documents and conversations.

Context window

Maximum tokens per request · higher is better

58 more models · see the full leaderboard

Score vs. context window

Accurate with room for long inputs sits top right

All models

Every model in one table. Select a column heading to sort.

All models compared
	Features			Accuracy	Price		Speed	Latency	End-to-end
#
1	Claude Opus 5.5	1M	Anthropic	85.0	$0.015	$4 / $20	129	2.3 s	7.6 s
2	Gemini 4 Argon	1M	Google	84.8	$0.0063	$2 / $10	69	3.3 s	9.9 s
3	GPT-5.3-Codex	400k	OpenAI	82.5	$0.01	$1.75 / $14	95	1.4 s	5.2 s
4	Gemini 3.8 Flash	1M	Google	82.5	$0.0037	$0.825 / $4.125	142	3.2 s	5.6 s
5	Claude Fable 5.1	1M	Anthropic	81.2	$0.016	$10 / $50	70	2.1 s	11.0 s
6	Grok 4.7	500k	SpaceXAI	80.0	$0.025	$2 / $6	77	1.9 s	8.0 s
7	Muse Wonder	1M	Meta	78.8	$0.021	$4 / $20	175	16.1 s	20.8 s
8	Claude Sonnet 5.5	1M	Anthropic	78.8	$0.0054	$2 / $10	126	0.7 s	6.2 s
9	Gemini 3.1 Pro Preview	1M	Google	77.5	$0.017	$2 / $12	124	3.4 s	8.9 s
10	Muse Spark Open	1M	Meta	77.5	$0.0045	$1.25 / $4.25	220	0.5 s	3.2 s
11	Kimi K3.1 Pro	1M	MoonshotAI	77.5	$0.01	$2 / $10	38	1.3 s	19.6 s
12	GPT-6 Astra	1.1M	OpenAI	76.2	$0.013	$10 / $50	49	1.3 s	8.9 s
13	GPT-6.1 Terra Pro	1.1M	OpenAI	76.2	$0.005	$1 / $5	—	8.5 s	9.0 s
14	GPT-6.1 Sol	1.1M	OpenAI	76.2	$0.0029	$2 / $10	41	2.1 s	10.8 s
15	GPT-5.6 Terra	1.1M	OpenAI	75.0	$0.0063	$2 / $12	101	1.1 s	5.3 s
16	Gemini 3.5 Flash Lite	1M	Google	75.0	$0.0034	$0.3 / $2.5	409	1.1 s	2.8 s
17	Qwen3.8 Max (0902)	1M	Qwen	75.0	$0.011	$2 / $6	46	1.2 s	12.2 s
18	GPT-6.1 Terra	1.1M	OpenAI	73.8	$0.0013	$1 / $5	122	2.4 s	5.8 s
19	DeepSeek V4.1 Flash 1017	1M	DeepSeek	73.8	$0.00091	$0.15 / $0.6	57	0.9 s	7.3 s
20	GPT-6 Sol	1.1M	OpenAI	73.8	$0.0024	$2 / $10	92	0.9 s	4.4 s
21	o3	200k	OpenAI	72.5	$0.0088	$2 / $8	163	1.2 s	3.9 s
22	GPT-5.4	1.1M	OpenAI	72.5	$0.011	$2.5 / $15	55	1.7 s	8.7 s
23	GPT-6 Luna	1.1M	OpenAI	72.5	$0.00025	$0.1 / $0.5	137	1.2 s	4.3 s
24	DeepSeek V4 Flash Vision Exp	1M	DeepSeek	71.2	$0.0011	$0.2156 / $0.6468	170	0.7 s	3.2 s
25	DeepSeek V4 Pro 0813	1M	DeepSeek	71.2	$0.0036	$1.32 / $3.96	56	0.8 s	10.1 s
26	Claude Haiku 5.5	1M	Anthropic	68.8	$0.00048	$0.1 / $0.5	192	0.7 s	3.9 s
27	MiMo-V2.6-Pro	1.1M	Xiaomi	68.8	$0.002	$0.435 / $0.87	58	0.7 s	7.9 s
28	GLM 5.3 Flash	1M	Z.ai	67.5	$0.00015	$0.15 / $0.5	83	0.5 s	7.4 s
29	Inkling Small	524k	Thinking Machines	67.5	$0.00047	$0.45 / $1.2	233	0.5 s	3.6 s
30	GLM 5.3	1M	Z.ai	67.5	$0.00093	$0.079 / $1.4	154	0.8 s	3.4 s
31	Inkling	524k	Thinking Machines	66.2	$0.0012	$1 / $4.05	118	0.6 s	5.0 s
32	MiniMax M3.1 Flash	1M	MiniMax	66.2	$0.00087	$0.15 / $0.6	131	0.5 s	4.2 s
33	Kimi K2.7 Code	262k	MoonshotAI	65.0	$0.006	$0.56 / $3.1	136	0.4 s	5.3 s
34	Step 5 Preview	1M	StepFun	63.8	$0.0047	$1 / $2.7	93	0.6 s	8.2 s
35	Qwen3.8 27B	1M	Qwen	63.8	$0.0043	$0.2 / $2.5	49	0.5 s	10.6 s
36	MiMo-V2.6-Flash	1.1M	Xiaomi	62.5	$0.00056	$0.14 / $0.28	77	0.4 s	5.4 s
37	Qwen3 Coder Next	262k	Qwen	62.5	$0.0017	$0.12 / $0.8	106	0.4 s	4.2 s
38	Qwen3.7 Plus	1M	Qwen	60.0	$0.0033	$0.32 / $1.28	62	1.1 s	12.4 s
39	Nemotron 3 Ultra	262k	NVIDIA	60.0	$0.0057	$0.6 / $2.4	136	0.8 s	3.6 s
40	Llama 4 Maverick	524k	Meta	59.0	$0.00059	$0.1875 / $0.6525	13	0.7 s	27.9 s
41	Hy3	262k	Tencent	58.8	$0.0012	$0.132 / $0.528	96	1.7 s	9.0 s
42	gpt-oss-120b	131k	OpenAI	58.8	$0.00013	$0.15 / $0.6	82	0.5 s	5.4 s
43	Mercury 2.5	260k	Inception	58.2	$0.00011	$0.04 / $0.15	157	2.3 s	6.5 s
44	Qwen3.8 2.4T A95B	1M	Qwen	57.5	$0.015	$2 / $6	44	3.1 s	14.8 s
45	MiniMax M3	1M	MiniMax	57.5	$0.0026	$0.3 / $1.2	121	0.5 s	4.2 s
46	gpt-oss-20b	131k	OpenAI	51.2	$0.000066	$0.018 / $0.09	103	0.7 s	4.0 s
47	Mistral Large 4	1M	Mistral	50.0	$0.011	$1.36 / $4.18	83	1.1 s	9.5 s
48	LongCat 2.0	1M	Meituan	48.8	$0.0032	$0.3 / $1.2	67	1.4 s	10.0 s
49	Command A	256k	Cohere	45.0	$0.007	$2.5 / $10	7	1.8 s	53.4 s
50	Solar Pro 4	524k	Upstage	45.0	$0.0018	$0.15 / $0.6	91	0.6 s	8.3 s
51	Nemotron 3.5 Lightning	262k	NVIDIA	43.8	$0.00049	$0.07 / $0.2	116	0.5 s	6.5 s
52	Ministral 3 8B 2512	262k	Mistral	42.5	$0.00027	$0.15 / $0.15	124	0.3 s	3.5 s
53	Llama 3.3 70B Instruct	131k	Meta	42.5	$0.00053	$0.1 / $0.32	42	0.7 s	8.8 s
54	Llama 4 Scout	1.3M	Meta	41.2	$0.00037	$0.1 / $0.3	15	1.2 s	25.6 s
55	Ling 3.1 Flash	262k	inclusionAI	41.2	$0.00061	$0.075 / $0.225	7	1.1 s	81.3 s
56	Phi 4	16k	Microsoft	41.2	$0.00012	$0.07 / $0.14	103	0.5 s	4.9 s
57	Solar Mini 4	524k	Upstage	41.2	$0.00053	$0.05 / $0.2	107	0.7 s	7.4 s
58	Nemotron 3.5 Super VL	1M	NVIDIA	40.0	$0.0013	$0.15 / $0.5	715	0.6 s	1.5 s
59	Qwen3.6 35B A3B	262k	Qwen	37.5	$0.0037	$0.1 / $0.95	135	0.4 s	5.6 s
60	Granite 4.2 8B	131k	IBM	36.2	$0.000077	$0.06 / $0.25	60	0.6 s	7.3 s
61	Ministral 3 14B 2512	262k	Mistral	33.8	$0.00031	$0.2 / $0.2	32	0.6 s	12.7 s
62	Qwen3.5 397B A17B	262k	Qwen	32.5	$0.013	$0.5 / $3.6	240	0.6 s	3.5 s
63	Mistral Small 4	262k	Mistral	32.5	$0.0019	$0.15 / $0.6	170	0.5 s	4.5 s
64	Gemma 4 31B	262k	Google	30.0	$0.0014	$0.2 / $0.4	58	0.6 s	12.6 s
65	Mistral Medium 3.5	262k	Mistral	28.8	$0.025	$1.5 / $7.5	145	0.5 s	5.4 s
66	Trinity Large Thinking	262k	Arcee AI	28.8	$0.0025	$0.25 / $0.8	329	0.6 s	2.7 s
67	Hermes 4 405B	131k	Nous	26.2	$0.01	$1 / $3	41	0.3 s	17.5 s
68	Ministral 3 3B 2512	131k	Mistral	26.2	$0.00019	$0.1 / $0.1	252	0.3 s	1.6 s
69	Granite 4.0 Micro	131k	IBM	25.0	$0.000051	$0.017 / $0.112	38	0.7 s	8.9 s
70	Nova Micro 1.0	128k	Amazon	22.5	$0.000087	$0.035 / $0.14	180	0.5 s	2.1 s
71	Gemma 4 26B A4B	262k	Google	15.0	$0.0011	$0.09 / $0.3	172	0.7 s	4.8 s
72	Reka Flash 3	66k	rekaai	1.2	$0.0008	$0.1 / $0.2	101	0.5 s	7.4 s
73	LFM2.5-2.6B	128k	LiquidAI	—	—	$0.35 / $0.5	—	—	—
74	Ling 3.0 Flash VL	131k	inclusionAI	—	—	$0.06 / $0.18	—	—	—
75	Qwen3.5-122B-A10B	262k	Qwen	—	—	$0.4 / $3.2	108	1.2 s	18.0 s
76	Qwen3 Next 80B A3B Thinking	262k	Qwen	—	—	$0.15 / $1.2	—	—	—
77	Nova Premier 1.0	1M	Amazon	—	—	$2.5 / $12.5	—	—	—
78	GPT-5.5 Pro	1.1M	OpenAI	—	—	$30 / $180	—	33.8 s	35.4 s

Frequently asked questions

Which AI model is the most accurate?: Claude Opus 5.5 by Anthropic scored highest in our 2026-10-27 run, answering 68 of 80 questions correctly (85.0%), out of 78 models tested.
Which AI model is the fastest?: Nemotron 3.5 Super VL by NVIDIA had the highest median output speed in our 2026-10-27 run: 715 tokens per second.
Which AI model is the cheapest to run?: Granite 4.0 Micro by IBM had the lowest billed cost per question in our 2026-10-27 run: $0.000051, reasoning tokens included.
Which AI model responds the quickest?: Ministral 3 3B 2512 by Mistral had the shortest median time to first token in our 2026-10-27 run: 0.3 s.
Which AI model gives a complete answer the quickest?: Nemotron 3.5 Super VL by NVIDIA had the shortest median end-to-end time in our 2026-10-27 run: 1.5 seconds to stream a full answer of about 300 words.
Which AI model has the largest context window?: Llama 4 Scout by Meta accepts the most context of the models we test: 1,310,720 tokens, according to OpenRouter's catalogue.
What is the best open-weights AI model?: Muse Spark Open by Meta has the highest verification score among open-weights models we test: 77.5%.
How many AI models do you compare?: 78 models from 28 labs, each tested through OpenRouter with the same questions and speed runs.