LLM API pricing compared: cost per 1M tokens for every major model

Updated · 4 min read

On 2026-11-18 the cheapest models from the big providers cost $0.10 per million input tokens and $0.50 per million output tokens (OpenAI GPT-6 Luna and Claude Haiku 5.5). The main mid-range models cluster at $2 / $10 (Claude Sonnet 5.5, GPT-6.2 Sol, Gemini 4 Carbon and Argon), and the top models cost $8 / $40 (Claude Fable 5.2, GPT-6.2 Astra). Prices below come from each provider's own pricing page; the calculator turns them into a monthly bill for your workload.

Price and cost calculator

Pick a preset or type your own numbers. Models are sorted from cheapest to most expensive for that workload.

ModelPer requestPer month
Claude Haiku 5.5Anthropicclaude-haiku-5-5$0.10 in / $0.50 out per 1M$0.00045$0.67
GPT-6 LunaOpenAIgpt-6-luna$0.10 in / $0.50 out per 1M$0.00045$0.67
Qwen3.8 FlashAlibaba (Qwen)qwen3.8-flash$0.15 in / $0.47 out per 1M$0.00053$0.80
Mistral Small 4Mistralmistral-small-latest$0.15 in / $0.60 out per 1M$0.00060$0.90
DeepSeek V4.1 FlashDeepSeekdeepseek-flash$0.30 in / $1.20 out per 1M$0.0012$1.80
Mistral Large 4Mistralmistral-large-latest$0.70 in / $2.10 out per 1M$0.0024$3.67
Gemini 3.8 FlashGooglegemini-3.8-flash$0.75 in / $3.75 out per 1M$0.0034$5.06
Grok 4.3xAIgrok-4.3$1.25 in / $2.50 out per 1M$0.0037$5.63
GPT-6.1 TerraOpenAIgpt-6.1-terra$1 in / $5 out per 1M$0.0045$6.75
DeepSeek V4 ProDeepSeekdeepseek-v4-pro$1.32 in / $3.96 out per 1M$0.0046$6.93
Qwen4 MaxAlibaba (Qwen)qwen4-max$1.50 in / $4.50 out per 1M$0.0053$7.88
Gemini 4 HydrogenGooglegemini-4-hydrogen$1.50 in / $7.50 out per 1M$0.0067$10.13
Grok 4.7xAIgrok-4.7$2 in / $6 out per 1M$0.0070$10.50
Claude Sonnet 5.5Anthropicclaude-sonnet-5-5$2 in / $10 out per 1M$0.0090$13.50
GPT-6.2 SolOpenAIgpt-6.2-sol$2 in / $10 out per 1M$0.0090$13.50
Gemini 4 CarbonGooglegemini-4-carbon$2 in / $10 out per 1M$0.0090$13.50
Gemini 4 ArgonGooglegemini-4-argon$2 in / $10 out per 1M$0.0090$13.50
Kimi K3.1 ProMoonshot AIkimi-k3.1-pro$2 in / $10 out per 1M$0.0090$13.50
Claude Opus 5.5Anthropicclaude-opus-5-5$4 in / $20 out per 1M$0.02$27.00
Claude Fable 5.2Anthropicclaude-fable-5-2$8 in / $40 out per 1M$0.04$54.00
GPT-6.2 AstraOpenAIgpt-6.2-astra$8 in / $40 out per 1M$0.04$54.00

USD list prices per 1M tokens, standard tier, checked on on each provider's pricing page. A month is 30 days. Taxes, batch discounts, tool fees (web search) and cache-write charges are left out.

Qwen3.8 Flash: Singapore-region price.

DeepSeek V4.1 Flash, DeepSeek V4 Pro: Peak-hour price. Off-peak (most of the week) is half.

Gemini 3.8 Flash: Also on the free tier. Google lists $1.50 input / $7.50 output from 2027-01-01.

Qwen4 Max: Singapore-region price; new accounts get a free quota there.

Gemini 4 Hydrogen: Also on the Gemini API free tier, with rate limits.

Gemini 4 Carbon, Gemini 4 Argon: Google lists $4 input / $20 output from 2026-12-01.

Long prompts cost more (input / output): Claude Haiku 5.5 $0.50 / $2.50 for prompts over 100K tokens; GPT-6 Luna $0.20 / $0.75 for prompts of 272K tokens or more; Grok 4.3 $2.50 / $5 for prompts of 200K tokens or more; GPT-6.1 Terra $2 / $7.50 for prompts of 272K tokens or more; Grok 4.7 $4 / $12 for prompts of 200K tokens or more; GPT-6.2 Sol $4 / $15 for prompts of 272K tokens or more; GPT-6.2 Astra $16 / $60 for prompts of 272K tokens or more. The calculator switches to these rates when your input reaches them (marked *).

Prices by tier

Prices are per 1M tokens, input / output, standard tier, on 2026-11-18.

Tier Model Input Output
Top Claude Fable 5.2 $8 $40
Top GPT-6.2 Astra $8 $40
Upper Claude Opus 5.5 $4 $20
Mid Claude Sonnet 5.5 $2 $10
Mid Kimi K3.1 Pro $2 $10
Mid GPT-6.2 Sol $2 $10
Mid Gemini 4 Carbon, Gemini 4 Argon $2 (from Dec 1: $4) $10 (from Dec 1: $20)
Mid Grok 4.7 $2 $6
Mid Gemini 4 Hydrogen $1.50 $7.50
Mid Qwen4 Max (Singapore) $1.50 $4.50
Mid DeepSeek V4 Pro (peak / off-peak) $1.32 / $0.66 $3.96 / $1.98
Value Grok 4.3 $1.25 $2.50
Value GPT-6.1 Terra $1 $5
Value Gemini 3.8 Flash $0.75 $3.75
Value Mistral Large 4 $0.70 $2.10
Value DeepSeek V4.1 Flash (peak / off-peak) $0.30 / $0.15 $1.20 / $0.60
Small Qwen3.8 Flash (Singapore) $0.15 $0.47
Small Mistral Small 4 $0.15 $0.60
Small Claude Haiku 5.5 $0.10 $0.50
Small GPT-6 Luna $0.10 $0.50

Tiers are our grouping by price, not a quality ranking. Two models at the same price can differ a lot in quality, speed and context length, so test on your own prompts.

What changes the real cost

  • Output costs 2 to 5 times as much as input. A chat app that writes long answers pays mostly for output; a summarizer that reads long documents pays mostly for input.
  • With prompt caching, apps that resend the same context (coding agents, long chats, documents) pay a much lower price for the repeated part. A cache hit costs $0.10 per million on Claude Sonnet 5.5 against $2 for fresh input, and $0.006 against $0.30 on deepseek-flash. Set "Input from cache" in the calculator to see the effect.
  • Some models charge more for the whole request when the prompt is long: Claude Haiku 5.5 above 100K tokens, GPT-6 models from 272K tokens, Grok from 200K tokens. The calculator switches to those rates automatically.
  • Anthropic and OpenAI charge half price for requests sent through their Batch APIs, which return results later instead of instantly.
  • DeepSeek charges half price outside 35 peak hours a week. See DeepSeek API pricing for the hours in your time zone.
  • A million tokens is not the same amount of text everywhere. Anthropic says Claude 4.7 and later models produce about 30% more tokens for the same text than earlier Claude models. Compare costs on your own prompts if the difference matters.
  • Web search, code execution and other server-side tools can cost extra per use. Claude's web search, for example, costs $10 per 1,000 searches.

Price changes already announced

  • Gemini 4 Carbon and Gemini 4 Argon go from $2 / $10 to $4 / $20 on 2026-12-01.
  • Gemini 3.8 Flash goes from $0.75 / $3.75 to $1.50 / $7.50 on 2027-01-01.

Free options

If you only need a few requests a day, you may not need to pay at all. Gemini 4 Hydrogen and Gemini 3.8 Flash are on Google's free tier (see Gemini API key), and OpenRouter has a rotating set of free models. Our free AI API keys guide compares every free option we know of. Models you run on your own computer cost nothing per token; see local LLM apps.

Provider pricing guides

How we keep this table current

Every price comes from the provider's own pricing page, checked on the date at the top. We re-check them every month and when a provider launches a model. If you spot a price that has changed, the about page says how to reach us.

Questions

What is the cheapest LLM API?

Among the big providers on 2026-11-18, OpenAI's GPT-6 Luna and Anthropic's Claude Haiku 5.5 both cost $0.10 per million input tokens and $0.50 per million output tokens. DeepSeek's deepseek-flash costs $0.15 / $0.60 off-peak and $0.30 / $1.20 at peak. For zero cost, use a free tier such as Gemini's or OpenRouter's free models.

How do I calculate the cost of an API request?

Multiply the input tokens by the input price and the output tokens by the output price, each divided by one million, and add them. Cached input is billed at the lower cache-hit price. The calculator on this page does it for every model at once.

How many words is a million tokens?

Roughly 750,000 English words, but it depends on the model's tokenizer. Anthropic says its Claude 4.7 and later models produce about 30% more tokens for the same text than its earlier models.

Why is output more expensive than input?

Generating text takes more compute than reading it, and every provider prices it higher: on 2026-11-18 output cost 5 times the input price for Claude, GPT and Gemini, 3 times for Grok 4.7, and 3 to 4 times for DeepSeek.

Will these prices change?

Yes. Google has already announced higher prices for Gemini 4 Carbon and Argon from 2026-12-01 and for Gemini 3.8 Flash from 2027-01-01. We re-check every provider's pricing page each month and date the table.

Sources

Checked against the providers' own documentation on 2026-11-18. Prices and limits change; the links below have the current figures.

Related guides