LLM API pricing comparison: what each provider charges per 1M tokens
Flagship and budget model prices from OpenAI, Anthropic, Google, DeepSeek and Mistral, taken from each provider’s own pricing page, with a worked monthly cost example and the discounts each one offers.
Last updated October 22, 2026
How do LLM API prices compare?
On October 22, 2026, flagship models cost $1.32 to $10 per 1M input tokens and $3.96 to $50 per 1M output tokens at standard (DeepSeek peak) rates, with gpt-6-astra and Claude Fable 5.1 the most expensive. Budget models start at $0.10 per 1M input tokens. Batch APIs halve prices at OpenAI, Anthropic, Google and Mistral.
What do flagship and budget models cost at each provider?
For each provider we picked the most capable general model it lists and its lowest-priced current text model. Prices are standard (pay-as-you-go) rates in US dollars per 1M tokens, taken from each provider’s own pricing page.
| Provider | Flagship model | Flagship input / output | Budget model | Budget input / output | Full prices |
|---|---|---|---|---|---|
| OpenAI | gpt-6-astra | $10.00 / $50.00 (prompts up to 272K tokens) | gpt-6-luna | $0.10 / $0.50 | OpenAI API pricing |
| Anthropic | Claude Fable 5.1 | $10 / $50 | Claude Haiku 5.5 | $0.10 / $0.50 (prompts up to 100,000 tokens) | Claude API pricing |
| Gemini 4 Argon | $2.00 / $10.00 through November 30, 2026; $4.00 / $20.00 from December 1, 2026 | Gemini 2.5 Flash-Lite | $0.10 / $0.40 (text, image, video) | Gemini API pricing | |
| DeepSeek | deepseek-v4-pro | $1.32 / $3.96 peak; $0.66 / $1.98 off-peak (cache miss) | deepseek-flash | $0.30 / $1.20 peak; $0.15 / $0.60 off-peak (cache miss) | DeepSeek API pricing |
| Mistral | Mistral Large 4 | $1.36 / $4.18 | Ministral 3 3B | $0.10 / $0.10 | Mistral API pricing |
Prices checked October 22, 2026. Providers change prices often; confirm on each provider’s page: OpenAI, Anthropic, Google Gemini, DeepSeek and Mistral.
“Flagship” is our label for the top general model on each price list, not a claim that these models are equal. They differ in quality, speed, context length and features, so compare them on your own tasks before choosing on price.
What would 1M requests a month cost on each model?
Take a workload of 1,000,000 requests a month, each with 1,000 input tokens and 300 output tokens. That is 1,000M input tokens and 300M output tokens. With prices per 1M tokens:
Monthly cost = 1,000 × input price + 300 × output price
| Model | Calculation | Monthly cost |
|---|---|---|
| gpt-6-astra | 1,000 × $10 + 300 × $50 | $25,000 |
| Claude Fable 5.1 | 1,000 × $10 + 300 × $50 | $25,000 |
| Gemini 4 Argon (through Nov 30, 2026) | 1,000 × $2 + 300 × $10 | $5,000 |
| Gemini 4 Argon (from Dec 1, 2026) | 1,000 × $4 + 300 × $20 | $10,000 |
| Mistral Large 4 | 1,000 × $1.36 + 300 × $4.18 | $2,614 |
| deepseek-v4-pro (peak, cache miss) | 1,000 × $1.32 + 300 × $3.96 | $2,508 |
| deepseek-flash (peak, cache miss) | 1,000 × $0.30 + 300 × $1.20 | $660 |
| gpt-6-luna | 1,000 × $0.10 + 300 × $0.50 | $250 |
| Claude Haiku 5.5 | 1,000 × $0.10 + 300 × $0.50 | $250 |
| Gemini 2.5 Flash-Lite | 1,000 × $0.10 + 300 × $0.40 | $220 |
| Ministral 3 3B | 1,000 × $0.10 + 300 × $0.10 | $130 |
The spread between the most and least expensive rows is almost 200 times. Most teams do not need the top model for every request, which is why routing simple requests to a cheaper model is one of the biggest levers on an LLM bill. To price your own prompts, count them with the token calculator, or model a full workload with the LLM cost calculator.
Which discounts does each provider offer?
Each provider lists its own ways to pay less than the standard rate. These are the ones stated on their pricing pages:
| Provider | Batch or async discount | Cached input | Other price levers |
|---|---|---|---|
| OpenAI | Batch and Flex are priced at half the standard rate | Cached input at 0.1× input on most models (0.05× on gpt-6.1-sol); cache writes cost 1.25× (30 min) or 2.5× (12 h) | Prompts over 272K input tokens cost more; Fast mode costs more (2× standard on gpt-6-astra) |
| Anthropic | Batch API: 50% off input and output | Cache hits at 0.1× input (0.025× on Fable 5.1, 0.05× on Opus 5.5 and Sonnet 5.5); 5-minute writes 1.25×, 1-hour writes 2× | US-only inference costs 1.1×; Haiku 5.5 costs more above 100,000 prompt tokens |
| Batch API: 50% cost reduction | Context caching priced per model, plus an hourly storage price on some models | Free tier on many models; Priority tier costs more; some prices rise on set dates | |
| DeepSeek | None listed on the pricing page | Cache-hit input is 2% of the cache-miss price ($0.006 vs $0.30 on deepseek-flash at peak) | Off-peak rates are half of peak rates |
| Mistral | Batch: 50% off | Cached input at about one tenth of the input price | Priority tier costs 1.75× standard |
Prices checked October 22, 2026. Providers change prices often; confirm on each provider’s page: OpenAI, Anthropic, Google Gemini, DeepSeek and Mistral.
How does Perimattic LLM Gateway help with these prices?
Perimattic LLM Gateway sits between your application and these providers. The features that bear on cost:
- One OpenAI-compatible API in front of OpenAI, Anthropic, Google, Mistral, DeepSeek and open models, so switching models is a configuration change.
- Routing rules that send each request to a model by cost, latency or task, with automatic fallback when a provider errors or rate-limits.
- Exact-match and semantic caching of responses.
- Budgets and rate limits per team, API key or customer, with cost tracking and alerts.
Sign up and our team will walk you through Perimattic LLM Gateway. To see what caching and routing might save on your workload, try the LLM gateway savings calculator.
Where do these prices come from?
Every price on this page was copied from the provider’s own pricing page on October 22, 2026: OpenAI API pricing, Anthropic Claude pricing, Gemini Developer API pricing, DeepSeek models and pricing and Mistral API pricing. Prices are in US dollars, exclude tax, and are the standard pay-as-you-go rates unless marked otherwise. Prices checked October 22, 2026; providers change prices often — confirm on the provider’s page before you rely on them. Perimattic is not affiliated with any of these providers, and model names are their owners’ trademarks.
Detailed pricing by provider
LLM API pricing, answered
Which LLM API is the cheapest?
Per token, among the models in our table, Ministral 3 3B from Mistral ($0.10 input and $0.10 output per 1M tokens) and DeepSeek’s deepseek-flash at off-peak rates ($0.15 input, $0.60 output) are the lowest. But cheaper models are not equal in quality, and a model that needs longer prompts or more retries can cost more in practice. Test two or three candidates on your own prompts.
How do input and output token prices differ?
Providers set the two prices separately, and almost every model we checked charges more for output than input; Mistral’s Ministral 3 models are the exception, priced the same both ways. Across the models on our provider pages, output costs from 1 to about 8 times the input price. Reasoning models also bill their thinking tokens as output, so long reasoning can dominate the bill.
Are token counts the same across providers?
No. Each provider uses its own tokenizer, so the same text produces a different number of tokens. Anthropic’s pricing page, for example, says Claude 4.7 and later models produce roughly 30% more tokens for the same text than earlier Claude models. Compare costs on real prompts, not only on the price per token.
Want one API for every model, with costs under control?
Sign up to tell us which models and providers you use, and our team will get in touch to set up the LLM Gateway for your stack.
Sign up