What is the cheapest LLM API?

Qwen3.7-Flash from Alibaba (Qwen) has the lowest blended price in our table at $0.055 per 1M tokens ($0.03 input, $0.13 output). If your requests are mostly output, check the output column too: the ranking changes.

What is the cheapest flagship LLM?

MiniMax-M3 is the lowest-priced flagship model at $0.30 per 1M input tokens and $1.20 per 1M output tokens.

Is the cheapest model good enough?

This page ranks price only, not quality. Small models handle classification, extraction and short answers well; harder reasoning and long code changes often need a larger model. Run a sample of your own prompts through two or three candidates before switching.

Is there a free LLM API?

Some providers offer free tiers with rate limits, for example the Google Gemini API. Free tiers usually come with lower limits and different data-use terms than paid usage, so read the terms before sending production data.

Cheapest LLM APIs (October 2026)

The cheapest LLM API right now is Qwen3.7-Flash at $0.055 per 1M tokens blended ($0.03 input, $0.13 output). Below are the five lowest-priced models in each group, from 50 models we track.

Rankings use a blended price that assumes 3 input tokens for every output token: (3 × input + output) ÷ 4. If your ratio is different, put your own numbers into the LLM Cost Calculator.

Cheapest overall

Every model available in the API, regardless of size.

#	Model	Provider	Blended $/1M	Input $/1M	Output $/1M	Context
1	Qwen3.7-Flash Alibaba (Qwen) · $0.03 in / $0.13 out	Alibaba (Qwen)	$0.055	$0.03	$0.13	1M
2	Ministral 3 3B Mistral · $0.10 in / $0.10 out	Mistral	$0.10	$0.10	$0.10	256K
3	Ministral 3 8B Mistral · $0.15 in / $0.15 out	Mistral	$0.15	$0.15	$0.15	256K
4	GPT-6 Luna OpenAI · $0.10 in / $0.50 out	OpenAI	$0.20	$0.10	$0.50	1.05M
5	Ministral 3 14B Mistral · $0.20 in / $0.20 out	Mistral	$0.20	$0.20	$0.20	256K

Cheapest flagship models

Each provider's top tier, by the provider's own naming.

#	Model	Provider	Blended $/1M	Input $/1M	Output $/1M	Context
1	MiniMax-M3 MiniMax · $0.30 in / $1.20 out	MiniMax	$0.525	$0.30	$1.20	1M
2	DeepSeek-V4-Pro DeepSeek · $1.32 in / $3.96 out	DeepSeek	$1.98	$1.32	$3.96	1M
3	GLM-5.3 Zhipu (GLM) · $1.40 in / $4.40 out	Zhipu (GLM)	$2.15	$1.40	$4.40	1M
4	GLM-5.2 Zhipu (GLM) · $1.40 in / $4.40 out	Zhipu (GLM)	$2.15	$1.40	$4.40	1M
5	Qwen3.8-Max Alibaba (Qwen) · $2 in / $6 out	Alibaba (Qwen)	$3	$2	$6	1M

Cheapest with a 1M-token context window

Models that accept roughly one million tokens or more.

#	Model	Provider	Blended $/1M	Input $/1M	Output $/1M	Context
1	Qwen3.7-Flash Alibaba (Qwen) · $0.03 in / $0.13 out	Alibaba (Qwen)	$0.055	$0.03	$0.13	1M
2	GPT-6 Luna OpenAI · $0.10 in / $0.50 out	OpenAI	$0.20	$0.10	$0.50	1.05M
3	Qwen3.8-Flash Alibaba (Qwen) · $0.15 in / $0.47 out	Alibaba (Qwen)	$0.23	$0.15	$0.47	1M
4	GLM-5.3-Flash Zhipu (GLM) · $0.15 in / $0.50 out	Zhipu (GLM)	$0.2375	$0.15	$0.50	1M
5	GPT-5.6 Luna OpenAI · $0.20 in / $1.20 out	OpenAI	$0.45	$0.20	$1.20	1.05M

Cheapest with image input

Models that accept images in the prompt.

#	Model	Provider	Blended $/1M	Input $/1M	Output $/1M	Context
1	Qwen3.7-Flash Alibaba (Qwen) · $0.03 in / $0.13 out	Alibaba (Qwen)	$0.055	$0.03	$0.13	1M
2	Ministral 3 3B Mistral · $0.10 in / $0.10 out	Mistral	$0.10	$0.10	$0.10	256K
3	Ministral 3 8B Mistral · $0.15 in / $0.15 out	Mistral	$0.15	$0.15	$0.15	256K
4	GPT-6 Luna OpenAI · $0.10 in / $0.50 out	OpenAI	$0.20	$0.10	$0.50	1.05M
5	Ministral 3 14B Mistral · $0.20 in / $0.20 out	Mistral	$0.20	$0.20	$0.20	256K

Cheapest open-weight models

Models whose weights are published, priced on the provider’s own API.

#	Model	Provider	Blended $/1M	Input $/1M	Output $/1M	Context
1	Ministral 3 3B Mistral · $0.10 in / $0.10 out	Mistral	$0.10	$0.10	$0.10	256K
2	Ministral 3 8B Mistral · $0.15 in / $0.15 out	Mistral	$0.15	$0.15	$0.15	256K
3	Ministral 3 14B Mistral · $0.20 in / $0.20 out	Mistral	$0.20	$0.20	$0.20	256K
4	Qwen3.8-Flash Alibaba (Qwen) · $0.15 in / $0.47 out	Alibaba (Qwen)	$0.23	$0.15	$0.47	1M
5	GLM-5.3-Flash Zhipu (GLM) · $0.15 in / $0.50 out	Zhipu (GLM)	$0.2375	$0.15	$0.50	1M

Input-heavy vs output-heavy work

The cheapest model depends on what you send and what you get back. Lowest input price: Qwen3.7-Flash at $0.03 per 1M tokens, which suits long prompts, retrieval and classification. Lowest output price: Ministral 3 3B at $0.10 per 1M tokens, which matters more for long answers and generated code.

Prices are each provider's standard pay-as-you-go rates. Cached input, batch discounts and off-peak pricing can lower them further; long-context surcharges can raise them.