AI Model Pricing
Prices are each model's typical published rate. AI usage is billed at these rates plus an 8% platform fee.
Rates as of . All prices are in US dollars per million tokens.
How to read this table
- Input is the prompt you send, the part that is not served from the cache.
- Cache read is prompt text the model has already seen in a recent request and reads back from its cache. Where a model has no separate cache price, those tokens are billed at the input price.
- Cache write is prompt text stored in the cache for later requests. Some models charge separately for how long the cache is kept (5 minutes or 1 hour); for every other model a cache write is billed at the input price.
- Output is everything the model writes, its reasoning included.
Your usage is drawn from your organization's universal usage credits, and the Billing & Usage screen shows your AI usage and what it drew. Using your own provider keys instead? See bring your own keys.
Models and Prices
| Model | Input | Cache read | Cache write | Output |
|---|---|---|---|---|
| Claude Opus 5.5 | $4 | $0.20 | $5 (5 min) / $8 (1 hour) | $20 |
| Claude Sonnet 5.5 | $2 | $0.10 | $2.50 (5 min) / $4 (1 hour) | $10 |
| Claude Fable 5.1 | $10 | $0.25 | $12.50 (5 min) / $20 (1 hour) | $50 |
| Claude Opus 4.6 | $5 | $0.50 | $6.25 (5 min) / $10 (1 hour) | $25 |
| Claude Sonnet 4.6 | $3 | $0.30 | $3.75 (5 min) / $6 (1 hour) | $15 |
| Claude Haiku 4.5 | $1 | $0.10 | $1.25 (5 min) / $2 (1 hour) | $5 |
| Claude Fable 5.2 | $8 | $0.20 | $10 (5 min) / $16 (1 hour) | $40 |
| Claude Haiku 5.5 | $0.10 | $0.01 | $0.125 (5 min) / $0.20 (1 hour) | $0.50 |
| Claude Opus 5 | $5 | $0.50 | $6.25 (5 min) / $10 (1 hour) | $25 |
| Claude Opus 4.8 | $5 | $0.50 | $6.25 (5 min) / $10 (1 hour) | $25 |
| Claude Sonnet 5 | $2 | $0.20 | $2.50 (5 min) / $4 (1 hour) | $10 |
| GPT-6 Astra | $10 | $1 | $10 (billed as input) | $50 |
| GPT-5.6 Sol | $4 | $0.40 | $4 (billed as input) | $20 |
| GPT-5.6 Terra | $2 | $0.20 | $2 (billed as input) | $12 |
| GPT-5.6 Luna | $0.20 | $0.02 | $0.20 (billed as input) | $1.20 |
| GPT-5.5 | $5 | $0.50 | $5 (billed as input) | $30 |
| GPT-5.4 | $2.50 | $0.25 | $2.50 (billed as input) | $15 |
| GPT-5.4 Pro | $30 | $30 (billed as input) | $30 (billed as input) | $180 |
| GPT-5.4 mini | $0.75 | $0.075 | $0.75 (billed as input) | $4.50 |
| GPT-5.4 nano | $0.20 | $0.02 | $0.20 (billed as input) | $1.25 |
| GPT-6.1 Sol | $2 | $0.10 | $2 (billed as input) | $10 |
| GPT-6.1 Terra | $1 | $0.05 | $1 (billed as input) | $5 |
| GPT-6 Luna | $0.10 | $0.01 | $0.10 (billed as input) | $0.50 |
| GPT-6 Sol | $2 | $0.20 | $2 (billed as input) | $10 |
| Gemini 3.1 Pro (preview) | $2 | $0.20 | $2 (billed as input) | $12 |
| Gemini 3.1 Flash-Lite | $0.25 | $0.025 | $0.25 (billed as input) | $1.50 |
| Gemini 3 Flash (preview) | $0.50 | $0.05 | $0.50 (billed as input) | $3 |
| Gemini 2.5 Pro | $1.25 | $0.125 | $1.25 (billed as input) | $10 |
| Gemini 2.5 Flash | $0.30 | $0.03 | $0.30 (billed as input) | $2.50 |
| Gemini 2.5 Flash-Lite | $0.10 | $0.01 | $0.10 (billed as input) | $0.40 |
| Gemini 4 Carbon | $2 | $0.10 | $2 (billed as input) | $10 |
| Gemini 4 Argon | $2 | $0.10 | $2 (billed as input) | $10 |
| Gemini 4 Hydrogen | $1.50 | $0.075 | $1.50 (billed as input) | $7.50 |
| Gemini 3.8 Flash | $0.75 | $0.075 | $0.75 (billed as input) | $3.75 |
| Gemini 3.6 Flash | $0.75 | $0.075 | $0.75 (billed as input) | $3.75 |
| Gemini 3.5 Flash-Lite | $0.30 | $0.03 | $0.30 (billed as input) | $2.50 |
| GLM-5.1 | $1.33 | $0.247 | $1.33 (billed as input) | $4.18 |
| GLM-4.7 | $0.60 | $0.11 | $0.60 (billed as input) | $2.20 |
| DeepSeek V3.2 | $0.26 | $0.13 | $0.26 (billed as input) | $0.38 |
| gpt-oss-120b | $0.15 | $0.015 | $0.15 (billed as input) | $0.60 |
| gpt-oss-20b | $0.03 | $0.03 | $0.03 (billed as input) | $0.13 |
| Qwen3 14B | $0.10 | $0.10 (billed as input) | $0.10 (billed as input) | $0.22 |
| Kimi K2.5 | $0.60 | $0.10 | $0.60 (billed as input) | $3 |
| Kimi K2 Thinking | $0.60 | $0.15 | $0.60 (billed as input) | $2.50 |
| DeepSeek R1 Distill Qwen 32B | $0.497 | $0.497 (billed as input) | $0.497 (billed as input) | $4.881 |
| Mistral Small 3.1 24B Instruct | $0.351 | $0.351 (billed as input) | $0.351 (billed as input) | $0.555 |
| Llama 4 Scout 17B 16E Instruct | $0.27 | $0.27 (billed as input) | $0.27 (billed as input) | $0.85 |
| Gemma 3 12B | $0.345 | $0.345 (billed as input) | $0.345 (billed as input) | $0.556 |
| Qwen3 30B A3B | $0.051 | $0.051 (billed as input) | $0.051 (billed as input) | $0.335 |
| Granite 4.0 H Micro | $0.017 | $0.017 (billed as input) | $0.017 (billed as input) | $0.112 |
| GLM-4.7 Flash | $0.06 | $0.06 (billed as input) | $0.06 (billed as input) | $0.40 |
| Nemotron 3 120B A12B | $0.50 | $0.50 (billed as input) | $0.50 (billed as input) | $1.50 |
| Gemma 4 26B A4B | $0.10 | $0.10 (billed as input) | $0.10 (billed as input) | $0.30 |
Long Prompts
A few models charge a higher rate when one request's whole prompt (input, cache reads and cache writes together) is above a size threshold. The whole request is then billed at the higher rate.
- Claude Haiku 5.5: Prompts over 100,000 tokens: input $0.50, cache read $0.05, output $2.50
- GPT-6 Astra: Prompts over 272,000 tokens: input $20, cache read $2, output $75
- GPT-5.6 Sol: Prompts over 272,000 tokens: input $8, cache read $0.80, output $30
- GPT-5.6 Terra: Prompts over 272,000 tokens: input $4, cache read $0.40, output $18
- GPT-5.6 Luna: Prompts over 272,000 tokens: input $0.40, cache read $0.04, output $1.80
- GPT-5.5: Prompts over 272,000 tokens: input $10, cache read $1, output $45
- GPT-5.4: Prompts over 272,000 tokens: input $5, cache read $0.50, output $22.50
- GPT-5.4 Pro: Prompts over 272,000 tokens: input $60, output $270
- GPT-6.1 Sol: Prompts over 272,000 tokens: input $4, cache read $0.20, output $15
- GPT-6.1 Terra: Prompts over 272,000 tokens: input $2, cache read $0.10, output $7.50
- GPT-6 Luna: Prompts over 272,000 tokens: input $0.20, cache read $0.02, output $0.75
- GPT-6 Sol: Prompts over 272,000 tokens: input $4, cache read $0.40, output $15
- Gemini 3.1 Pro (preview): Prompts over 200,000 tokens: input $4, cache read $0.40, output $18
- Gemini 2.5 Pro: Prompts over 200,000 tokens: input $2.50, cache read $0.25, output $15
Announced Price Changes
Some model makers have already announced a new price for a later date. From that date, requests are billed at the new price.
- Gemini 4 Carbon: From December 1, 2026: input $4, cache read $0.20, output $20
- Gemini 4 Argon: From December 1, 2026: input $4, cache read $0.20, output $20
- Gemini 3.8 Flash: From January 1, 2027: input $1.50, cache read $0.15, output $7.50
- Gemini 3.6 Flash: From January 1, 2027: input $1.50, cache read $0.15, output $7.50
A Worked Example
A request to a model priced at $0.30 input and $1.20 output that sends 20,000 input tokens and receives 2,000 output tokens costs 20,000 × $0.30 / 1,000,000 + 2,000 × $1.20 / 1,000,000 = $0.0084 at the model's rate, and $0.009072 with the 8% platform fee.
Questions
Do these prices change?
When a model's typical published rate changes, this table changes with it, and the date at the top of the page
moves to the day the rates were read. A request is billed at the rates in effect when it runs.
Why is a model I use elsewhere missing?
This table lists the built-in models available to you, and a new model appears here when it becomes available
to you.