AI model rankings

Which AI model is best?

Language models (the AI behind chatbots like ChatGPT) compared on how smart, fast and affordable they are. Based on independent tests, not vendor claims.

Updated Oct 30, 2026, 01:41 PM UTC Data from Artificial Analysis

Smartest

Gemini 4 Carbon

GoogleHigh

59intelligence score

Fastest

K2 Horizon 375B A23B

Institute of Foundation Models

750tokens / sec

Cheapest

Nemotron 3.5 Lightning

NVIDIA

$0.10per 1M tokens

Best value

GPT-6.1 Terra

OpenAIMax

51points for $2.00

Just released

All new models

MistralyesterdayLanguageNew

Introducing Mistral Large 4

Scores 26 on intelligence (#55 of 87), $1.05 per 1M tokens, about 127 words per second.

Google2 days agoLanguageNew

Introducing Gemini 4 Carbon

Scores 59 on intelligence (#1 of 87), $8.00 per 1M tokens, about 66 words per second.

OpenAI4 days agoLanguageNew

Introducing GPT-6.1 Terra

Scores 51 on intelligence (#12 of 87), $2.00 per 1M tokens, about 68 words per second.

Intelligence ≥ 0

Showing 87 of 87 models

Top 10 models by intelligence

Overall score across reasoning, knowledge and coding tests. Higher is better. Click a bar to see the model.

Compare all models

87 models · click a column to sort, click a model to see its page

#	Model
1	Gemini 4 CarbonNew GoogleHigh	59	–	88	66s	$4.00 / $20.0	Oct 2026
2	Claude Opus 5.5 AnthropicMax	58	–	97	563s	$4.00 / $20.0	Sep 2026
3	Claude Opus 5.5 AnthropicXhigh	57	–	83	84s	$4.00 / $20.0	Sep 2026
4	GPT-6 Astra OpenAIMax	57	–	88	184s	$10.0 / $50.0	Sep 2026
5	GPT-6 Astra OpenAIXhigh	55	–	76	66s	$10.0 / $50.0	Sep 2026
6	GPT-6 Astra OpenAIHigh	55	–	72	29s	$10.0 / $50.0	Sep 2026
7	GPT-6.1 Sol OpenAIMax	55	–	84	176s	$2.00 / $10.0	Sep 2026
8	Claude Sonnet 5.5 AnthropicMax	54	–	134	272s	$2.00 / $10.0	Sep 2026
9	GPT-6.1 Sol OpenAIXhigh	52	–	74	63s	$2.00 / $10.0	Sep 2026
10	Claude Opus 5.5 AnthropicHigh	52	–	80	18s	$4.00 / $20.0	Sep 2026
11	GPT-6 Astra OpenAIMedium	52	–	62	5.3s	$10.0 / $50.0	Sep 2026
12	GPT-6.1 TerraNew OpenAIMax	51	–	91	171s	$1.00 / $5.00	Oct 2026
13	Gemini 4 Argon GoogleHigh	50	–	80	28s	$4.00 / $20.0	Sep 2026
14	GPT-6.1 Sol OpenAIHigh	50	–	62	29s	$2.00 / $10.0	Sep 2026
15	Muse Wonder Full MetaMax	50	–	0	0.0s	$0.00 / $0.00	Oct 2026
16	Claude Fable 5.1 AnthropicXhigh	50	–	63	42s	$10.0 / $50.0	Sep 2026
17	Claude Sonnet 5.5 AnthropicXhigh	50	–	104	14s	$2.00 / $10.0	Sep 2026
18	Claude Fable 5.1 AnthropicMax	49	–	72	164s	$10.0 / $50.0	Sep 2026
19	Claude Opus 5.5 AnthropicMedium	48	–	75	14s	$4.00 / $20.0	Sep 2026
20	Claude Fable 5.1 AnthropicHigh	47	–	55	12s	$10.0 / $50.0	Sep 2026
21	GPT-6.1 Sol OpenAIMedium	46	–	57	3.7s	$2.00 / $10.0	Sep 2026
22	Kimi K3.1 ProNew Kimi	46	–	30	2.7s	$2.00 / $10.0	Oct 2026
23	GPT-6.1 TerraNew OpenAIXhigh	45	–	91	39s	$1.00 / $5.00	Oct 2026
24	Muse Wonder MetaMax	43	–	158	25s	$4.00 / $20.0	Oct 2026
25	GPT-6 Astra OpenAILow	43	–	66	1.6s	$10.0 / $50.0	Sep 2026

What do these numbers mean?

Intelligence: One overall score from many tests of reasoning, knowledge, maths and coding. Higher means it handles harder questions.
Speed (tokens/s): How fast the answer is written. A token is about ¾ of a word, so 100 tokens/s is roughly 75 words per second.
First token: How long you wait before the answer starts appearing. Reasoning models think first, so they take longer.
Price per 1M tokens: What developers pay through the API. 1 million tokens is about 750,000 words. Chat apps often hide this behind a monthly plan.

Benchmark data from Artificial Analysis, an independent benchmarking company. Rankings refresh every few hours.