Pricing
A flat monthly plan with unlimited tokens up to a cap in tokens per second, or pay for the tokens you use.
API plans
Unlimited tokens for a flat monthly price, paced to the plan's cap in tokens per second, in place of paying per token.
- What counts
- Output tokens count in full, uncached input tokens count as 0.33 of a token each, and cached input tokens are free. Every API key in the workspace shares the cap.
- Over the cap
- Responses slow to the plan's cap; a request is not refused for the plan. Up to 8 requests run at once per workspace; more wait in a queue and get 429 with a retry time after about 60 seconds.
- Credits
- Credits are not charged while a plan is in force, and stay in the workspace.
- Sail
- Sail keys are not covered. Sail plans are coming soon.
Pay per token
Add credits, create an API key, and each request is charged from the credits when it ends.
RegionUnited Statesus-region
| Model | Input | Cached input | Output |
|---|---|---|---|
| DeepSeek V4.1 Flashdeepseek/deepseek-v4.1-flash | $0.30 | $0.03 | $1.00 |
USD per 1M tokens. Credits never expire.
Chat
The web chat in your account, on the same models. The chat in your account stores its conversations until you delete them.
- Free
- 30 messages a day, 1 reply at a time, no attachments. Sign-up asks for a card, which is not charged.
- Sail plans
- Sail plans include the chat with no message cap, more than one reply at a time, and attachments of up to 10 MB a file and 5 a message. Plans are coming soon.
Questions
- How am I charged?
- On an API plan, a flat price a month in USD through Stripe, and its requests are not charged to credits. Without a plan, per token from prepaid credits: each request is charged for its input, cached input and output tokens at the listed price for its model and region. Credits are bought in USD from $10 and never expire.
- What is cached input?
- Input tokens the engine reads from the prefix cache. They are prompt tokens like any other, billed at the cached input price instead of the input price.
- Is anything kept?
- Prompts stay in memory for the session and are never written to disk. The session lasts at most 30 minutes after the last request on a key. Nothing is used to train a model. Nothing you send to the API is kept. The chat in your account stores its conversations until you delete them. A usage record (time, model, region, tokens, key and cost) is kept for billing. What we keep
- Do prices differ by region?
- Yes. Each model has its own price in each region: United States. A request runs in the region its base URL names and is not sent to another region.
- Can I change or cancel an API plan?
- Cancel from Billing, in the Stripe Customer Portal. The plan runs to the end of the month paid for. To change tier, cancel, then subscribe to the new tier: it starts at once, and the rest of the old month is not refunded. Changing tier in place, with the difference prorated, comes later.