Cheapest LLM API in 2026: price per 1M tokens, ranked

Cheapest LLM API in 2026

Updated 2026-10-18. Direct list prices from official pages, ranks from LMArena.

Quick answer

The cheapest LLM API right now is GPT-6 Luna from OpenAI at $0.10 input and $0.50 output per 1M tokens, then Qwen3.8 Flash ($0.15 / $0.47) and Mistral Small 4 ($0.15 / $0.60). The cheapest model in the top 8 for coding is Claude Sonnet 5.5 ($2 / $10). Prices are direct list prices checked 2026-10-18.

#	Model	Input	Output	Blended	Chat rank
1	GPT-6 Luna OpenAI	$0.10	$0.50	$0.20	#22
2	Qwen3.8 Flash Qwen	$0.15	$0.47	$0.23	n/a
3	Mistral Small 4 Mistral	$0.15	$0.60	$0.26	n/a
4	DeepSeek V4.1 Flash DeepSeek	$0.30	$1.20	$0.52	#15
5	Qwen3.7 Plus Qwen	$0.40	$1.60	$0.70	#19
6	Gemini 3.5 Flash-Lite Google	$0.30	$2.50	$0.85	#20
7	Claude Haiku 5.5 Anthropic	$0.50	$2.50	$1	n/a
8	Mistral Large 4 Mistral	$0.68	$2.09	$1.03	#24
9	Grok Build 0.1 xAI	$1	$2	$1.25	n/a
10	Gemini 3.8 Flash Google	$0.75	$3.75	$1.50	#5

Show all 29 models

#	Model	Input	Output	Blended	Chat rank
1	GPT-6 Luna OpenAI	$0.10	$0.50	$0.20	#22
2	Qwen3.8 Flash Qwen	$0.15	$0.47	$0.23	n/a
3	Mistral Small 4 Mistral	$0.15	$0.60	$0.26	n/a
4	DeepSeek V4.1 Flash DeepSeek	$0.30	$1.20	$0.52	#15
5	Qwen3.7 Plus Qwen	$0.40	$1.60	$0.70	#19
6	Gemini 3.5 Flash-Lite Google	$0.30	$2.50	$0.85	#20
7	Claude Haiku 5.5 Anthropic	$0.50	$2.50	$1	n/a
8	Mistral Large 4 Mistral	$0.68	$2.09	$1.03	#24
9	Grok Build 0.1 xAI	$1	$2	$1.25	n/a
10	Gemini 3.8 Flash Google	$0.75	$3.75	$1.50	#5
11	Grok 4.3 xAI	$1.25	$2.50	$1.56	#23
12	DeepSeek V4 Pro DeepSeek	$1.32	$3.96	$1.98	#17
13	Muse Spark 1.3 Meta	$1.25	$4.25	$2	#6
14	Qwen3.8 Max Qwen	$2	$6	$3	#11
15	Mistral Medium 3.5 Mistral	$1.50	$7.50	$3	#25
16	Grok 4.7 xAI	$2	$6	$3	#21
17	GPT-6.1 Sol OpenAI	$2	$10	$4	#10
18	GPT-6 Sol OpenAI	$2	$10	$4	#18
19	Claude Sonnet 5.5 Anthropic	$2	$10	$4	#14
20	Gemini 4 Argon Google	$2	$10	$4	#1
21	GPT-5.6 Terra OpenAI	$2	$12	$4.50	#16
22	Gemini 3.1 Pro Preview Google	$2	$12	$4.50	#8
23	GPT-5.6 Sol OpenAI	$4	$20	$8	#9
24	Claude Opus 5.5 Anthropic	$4	$20	$8	#3
25	Muse Wonder 1.0 Meta	$4	$20	$8	#2
26	Claude Opus 5 Anthropic	$5	$25	$10	#7
27	GPT-5.5 OpenAI	$5	$30	$11.25	#12
28	GPT-6 Astra OpenAI	$10	$50	$20	#13
29	Claude Fable 5.1 Anthropic	$10	$50	$20	#4

USD per 1M tokens, direct from each provider, checked 2026-10-18. Blended = (3 x input + output) / 4, a typical chat mix. Chat rank is the position among models we track on LMArena (CC BY 4.0) Text Arena.

Cheapest by job

Cheapest overall: GPT-6 Luna (OpenAI), $0.10 in and $0.50 out. A month of 30M input and 6M output tokens costs $6.
Cheapest good chat model: Gemini 3.8 Flash ($0.75 / $3.75), #5 on Text Arena among tracked models.
Cheapest strong coding model: Claude Sonnet 5.5 ($2 / $10), #3 on Code Arena among tracked models.
Off-peak discount: DeepSeek charges half price outside peak hours, so V4.1 Flash drops to $0.15 in and $0.60 out.
Free option: if you have the hardware, open models run locally at no per-token cost. The local model finder shows what fits your machine.

Temporary prices to watch

GPT-5.6 Sol: Promotional price, at least through Nov 21, 2026.
Gemini 4 Argon: Launch price through Nov 30, 2026. From Dec 1: $4 in, $20 out.
Gemini 3.8 Flash: Through Dec 31, 2026. From Jan 1, 2027: $1.50 in, $7.50 out.
Qwen3.7 Plus: List price, prompts up to 256K. Limited-time 20% off.
Mistral Large 4: Sale price. Regular: $1.36 in, $4.18 out.

We list direct provider prices only. Resellers like OpenRouter may differ. For a head-to-head view, see OpenAI vs Anthropic API pricing or the best model per task picks.

Sources

FAQ

What is the cheapest LLM API?

GPT-6 Luna from OpenAI at $0.10 input and $0.50 output per 1M tokens, as of 2026-10-18.

What is the cheapest LLM API for coding?

Among models in the top 8 of LMArena's Code Arena that we track, Claude Sonnet 5.5 is cheapest at $2 input and $10 output per 1M tokens.

Which LLM API has the best price for the quality?

It depends on the task. Our best model pages list a value pick per task from LMArena scores and our price table.

Is there a free LLM API?

Some providers offer free tiers with tight rate limits, and open models can run locally at no per-token cost. This list covers paid list prices only.

How often are these prices checked?

Daily, against each provider's official pricing page. Qwen's page is checked by hand.