LLM API pricing, ranked by cost

Every current first-party text model, priced on the workload you actually run. Cheapest depends on the shape of the work: a model that wins on a retrieval workload can lose badly on a content one. Rates captured 2026-11-10 from public pricing data and committed to our archive.

Workload profile

800k input, 200k output, 70% of input served from cache. Long context read repeatedly.

#
1	Ministral 3 3B 2512	Mistral	$0.100	$0.100	$0.010	$0.050
2	Ministral 3 8B 2512	Mistral	$0.150	$0.150	$0.015	$0.074
3	Devstral Small	Mistral	$0.100	$0.300	$0.010	$0.090
4	GPT 5 Nano	OpenAI	$0.050	$0.400	$0.005	$0.095
5	Ministral 3 14B 2512	Mistral	$0.200	$0.200	$0.020	$0.099
6	Claude Haiku 5.5	Anthropic	$0.100	$0.500	$0.010	$0.130
7	GPT 6 Luna	OpenAI	$0.100	$0.500	$0.010	$0.130
8	Mistral vibe cli Fast	Mistral	$0.150	$0.600	$0.015	$0.164
9	Codestral 2508	Mistral	$0.300	$0.900	$0.030	$0.269
10	GPT 5.6 Luna	OpenAI	$0.200	$1.20	$0.020	$0.299
11	GPT 5.4 Nano	OpenAI	$0.200	$1.25	$0.020	$0.309
12	DeepSeek Flash	DeepSeek	$0.300	$1.20	$0.006	$0.315
13	DeepSeek V4.1 Flash 1017	Mistral	$0.300	$1.20	$0.007	$0.316
14	Gemini 3.1 Flash Lite	Google	$0.250	$1.50	$0.025	$0.374
15	GPT 5 Mini	OpenAI	$0.250	$2.00	$0.025	$0.474
16	Mistral Code agent	Mistral	$0.400	$2.00	$0.040	$0.518
17	Gemini 3.5 Flash Lite	Google	$0.300	$2.50	$0.030	$0.589
18	Mistral Large 4	Mistral	$0.700	$2.10	$0.070	$0.627
19	Grok Code Fast	xAI	$1.00	$2.00	$0.200	$0.752
20	Grok 4.3	xAI	$1.25	$2.50	$0.200	$0.912
21	Gemini 3.8 Flash	Google	$0.750	$3.75	$0.075	$0.972
22	Sonar	Perplexity	$1.00	$1.00	—	$1.00
23	GPT 5.4 Mini	OpenAI	$0.750	$4.50	$0.075	$1.12
24	DeepSeek V4 Pro	DeepSeek	$1.32	$3.96	$0.044	$1.13
25	GPT 6.1 Terra	OpenAI	$1.00	$5.00	$0.050	$1.27
26	Sonar Reasoning	Perplexity	$1.00	$5.00	—	$1.80
27	Mistral Medium 3.5	Mistral	$1.50	$7.50	$0.150	$1.94
28	Grok 4.7	xAI	$2.00	$6.00	$0.500	$1.96
29	Gemini 2.5 Pro	Google	$1.25	$10.00	$0.125	$2.37
30	GPT 5 Chat	OpenAI	$1.25	$10.00	$0.125	$2.37
31	Claude Sonnet 5.5	Anthropic	$2.00	$10.00	$0.100	$2.54
32	Gemini 4 argon	Google	$2.00	$10.00	$0.100	$2.54
33	GPT 6.1 Sol	OpenAI	$2.00	$10.00	$0.100	$2.54
34	GPT 5.6 Terra	OpenAI	$2.00	$12.00	$0.200	$2.99
35	Gemini omni 1.2 Flash	Google	$1.50	$9.00	—	$3.00
36	Sonar Reasoning Pro	Perplexity	$2.00	$8.00	—	$3.20
37	GPT 5.3	OpenAI	$1.75	$14.00	$0.175	$3.32
38	Claude Opus 5.5	Anthropic	$4.00	$20.00	$0.200	$5.07
39	GPT 5.6	OpenAI	$4.00	$20.00	$0.400	$5.18
40	Sonar Pro	Perplexity	$3.00	$15.00	—	$5.40
41	GPT rosalind discovery	OpenAI	$5.00	$25.00	$0.500	$6.48
42	Claude Fable 5.2	Anthropic	$8.00	$40.00	$0.200	$10.03
43	GPT 6 astra	OpenAI	$10.00	$50.00	$1.00	$12.96
44	GPT 6 rosalind	OpenAI	$15.00	$75.00	$1.50	$19.44
45	GPT 6 Cyber	OpenAI	$20.00	$100.00	$2.00	$25.92

45 models priced on the retrieval profile. Models without a published cache-read rate are billed at the standard input rate for the whole input, and their cache column reads em dash.

Waiting for a price cut is not a plan

Every line is a model we track, priced on the workload you picked above. Notice they run flat. A published rate almost never moves once it is set, so the rate you pay today is very likely the rate you will pay next year. The cheaper lines lower down are not price cuts; they are newer models that launched underneath the old ones, and moving to one is the lever you actually have. Hover any line to see what that model charges for your workload, and check back when you next review costs.

Provider

Log scale

Highlight a model

Embed this chart

Paste this where you want the chart. It stays live: the lines redraw from our archive as rates move, and the reader can switch workload, filter by provider and pick out a model exactly as here. No script of ours runs on your page.

If the model you use is not on this list

This table shows only models a provider will sell you today. If the one you are running is missing, it has been retired or replaced, and a migration is coming whether or not you have planned for one. We keep its final rates and its full recorded history, so you can compare what you are paying now against what is actually on sale, and the ticker at the top of every page carries its last observed change.

Price your own workload Plain rate card Rates as JSON How to compare cheapness honestly

Using one of these figures

Rates on this page were captured 2026-11-10 from each provider's published pricing and committed to our archive, so this reading does not change when a provider edits a page. Quote the date with the number.

LLMPrice.io, “LLM API pricing compared by workload”. Rates captured 2026-11-10. https://llmprice.io/models

The same rates as JSON · how the archive works · questions about these numbers