AI API prices per million tokens, compared (October 2026)
Input, cached and output prices per million tokens for GPT-6, Claude 5, Gemini 4, DeepSeek V4, Grok and Mistral models, from each provider's own pricing page.
The cheapest current models from the big providers cost $0.10 to $0.30 per million input tokens, and the flagships $1.30 to $10. Output costs more everywhere: two to eight times the input price, and exactly five times on most OpenAI, Anthropic and Google models. The table below lists 17 models from six providers, each read on October 26, 2026 from the provider’s own pricing page, at standard (non-batch) rates.
Prices per million tokens
| Model | Provider | Input | Cached input | Output | Context window |
|---|---|---|---|---|---|
| GPT-6 Astra | OpenAI | $10.00 | $1.00 | $50.00 | 1.05M |
| GPT-6.1 Sol | OpenAI | $2.00 | $0.10 | $10.00 | 1.05M |
| GPT-6 Luna | OpenAI | $0.10 | $0.01 | $0.50 | 1.05M |
| Claude Fable 5.1 | Anthropic | $10.00 | $0.25 | $50.00 | 1M |
| Claude Opus 5.5 | Anthropic | $4.00 | $0.20 | $20.00 | 1M |
| Claude Sonnet 5.5 | Anthropic | $2.00 | $0.10 | $10.00 | 1M |
| Claude Haiku 5.5 | Anthropic | $0.10 | $0.01 | $0.50 | 1M |
| Gemini 4 Argon | $2.00 | $0.10 | $10.00 | 1,048,576 | |
| Gemini 3.8 Flash | $0.75 | $0.075 | $3.75 | 1,048,576 | |
| Gemini 3.5 Flash-Lite | $0.30 | $0.03 | $2.50 | 1,048,576 | |
| DeepSeek-V4-Pro | DeepSeek | $1.32 | $0.044 | $3.96 | 1M |
| DeepSeek-V4.1-Flash | DeepSeek | $0.30 | $0.006 | $1.20 | 1M |
| Grok 4.7 | xAI | $2.00 | $0.50 | $6.00 | 500K |
| Grok 4.3 | xAI | $1.25 | $0.20 | $2.50 | 1M |
| Mistral Large 4 | Mistral | $1.36 | $0.14 | $4.18 | 1M |
| Mistral Medium 3.5 | Mistral | $1.50 | not listed | $7.50 | 256K |
| Mistral Small 4 | Mistral | $0.15 | not listed | $0.60 | 256K |
The fine print that changes the bill
Long prompts can cost more
- OpenAI: the prices above apply up to 272,000 input tokens. Above that, GPT-6 Astra costs $20 input, $2 cached and $75 output; GPT-6.1 Sol $4, $0.20 and $15; GPT-6 Luna $0.20, $0.02 and $0.75.
- Anthropic: Fable 5.1, Opus 5.5 and Sonnet 5.5 bill the full 1M context at standard prices. Haiku 5.5’s prices apply to prompts up to 100,000 tokens; above that it is $0.50 input, $0.05 cached and $2.50 output.
- xAI: at 200,000 prompt tokens or more, the whole request is billed at a higher rate: $4, $1 and $12 for Grok 4.7; $2.50, $0.40 and $5 for Grok 4.3.
Prices that are about to change
Google’s pricing page announces new prices for Gemini 4 Argon from December 1, 2026 ($4 input, $0.20 caching, $20 output) and for Gemini 3.8 Flash from January 1, 2027 ($1.50, $0.15 and $7.50).
Off-peak discounts
DeepSeek lists the prices above as peak rates. Its off-peak rates are half: $0.66, $0.022 and $1.98 for V4-Pro, $0.15, $0.003 and $0.60 for V4.1-Flash. DeepSeek defines peak as 01:00 to 04:00 and 06:00 to 10:00 UTC, Monday to Friday, excluding Chinese public holidays.
Cache writes
The cached price is what you pay when the provider reuses an earlier prompt. Anthropic also charges for writing to the cache: for Sonnet 5.5, $2.50 per million tokens for a 5-minute cache and $4 for a 1-hour cache. Google adds a storage charge for context caching. Our guide to prompt caching covers when it pays off.
One price we could not settle
Mistral’s documentation lists Mistral Large 4 at $1.36 input and $4.18 output, while mistral.ai/pricing says “Mistral Large” costs $0.50 and $1.50 without naming a version. The table uses the documentation figure for the named model; check both pages before you budget around it.
What a workload actually costs
Per-million prices are hard to picture, so here is one workload priced on three models. A support chat that handles 10,000 conversations a month, each sending 1,500 tokens and getting 300 back, uses 15 million input and 3 million output tokens:
- GPT-6 Luna or Claude Haiku 5.5: 15 × $0.10 + 3 × $0.50 = $3.00 a month.
- Claude Sonnet 5.5 or Gemini 4 Argon: 15 × $2 + 3 × $10 = $60 a month.
- GPT-6 Astra: 15 × $10 + 3 × $50 = $300 a month.
A hundredfold spread for the same traffic is why picking the smallest model that does the job well is the biggest saving available. Run your own numbers in the API cost calculator, which prices any workload on all 17 models and handles caching.
Sources
- OpenAI, API pricing and models.
- Anthropic, pricing and models overview.
- Google, Gemini API pricing and models.
- DeepSeek, models and pricing.
- xAI, models and pricing.
- Mistral, pricing and mistral.ai/pricing.
All pages read on October 26, 2026.
Free toolWhat will 10,000 chats a month cost?Estimate my bill
