Pricing
Sign up >>>| Model | Input | Cached input | Output | Context | Zero Data Retention |
|---|---|---|---|---|---|
Particle-1particle-1 | /1M | /1M | /1M | 1M tokens | |
DeepSeek V4.1 Flashdeepseek-v41-flash-1017 | /1M | /1M | /1M | 1M tokens | |
GLM 5.3 Flashglm-5.3-flash | /1M | /1M | /1M | 1M tokens |
Web search
/ 1,000 searches
More intelligence less compute
Particle provides fast inference, high cache hit rates, and reliable infrastructure so you can focus on building great applications.
Sign up >>>Questions answered
Pricing is simple, but a few details matter. Here’s how billing, discounts and data retention work, in plain terms.
Find more in the docs >>>01How does billing work?
Particle runs on prepaid credits. Top up by card from the Billing page, optionally with auto top-up. Each request is charged per token against your balance, which is shared across all of your API keys with usage tracked per key.
02What counts as input, cached input and output?
Input is your prompt minus any prefix served from cache. Cached input is that cached prefix, billed at the lower cached rate. Output is everything the model generates, including reasoning tokens, which are not charged separately.
03What is Flex?
Flex is a service tier for work that can wait a little when the service is busy. Requests that run on Flex receive 50% off their applicable token prices. Check the X-Particle-Service-Tier header to confirm the tier a request ran on. Web-search fees are unchanged.
04Is web search extra?
Yes. Hosted web search costs per 1,000 searches and is billed separately from model tokens. Flex discounts do not apply to searches, and searches with empty results are still billed.
05What happens when my credits run out?
Requests are checked for a positive balance before any work starts. With an empty balance the API returns 402 and nothing is processed or charged. Turn on auto top-up to avoid interruptions.
06Do you store my prompts or train on them?
No. Standard inference runs under zero data retention: prompts and outputs are not written to persistent storage, and customer content is never used for training. See the Zero Data Retention policy.
07Where can I see my usage and spend?
Every response reports exact token counts, and the platform Billing page breaks spend down by day and model. Programmatic rollups are available through the analytics API.