Pricing
Updated: 2026-11-19 · All prices are USD list prices
Nexith bills by actual usage, deducted from the account balance. Balances are recorded as integer microdollars (1e-6 USD); displays and invoices are in USD.
Text and reasoning (nexith-core)#
| Item | Price |
|---|---|
| Input | $8.00 / 1M tokens |
| Cached input | $0.80 / 1M tokens (10% of the input price) |
| Output | $30.00 / 1M tokens |
Billing basis:
- Input: all messages in the request (system, history, and image-converted tokens).
- Cached input: portions that hit the upstream prefix cache are billed at 10%, reflected in
usage.prompt_tokens_details.cached_tokens. - Output: every token the model generates (including hidden reasoning tokens for reasoning models, if the upstream counts them).
- Streaming: usage is not returned by default; sending
stream_options: {"include_usage": true}delivers an authoritativechoices: []+usageframe before[DONE], and billing follows that frame.
Each request reserves a worst-case amount up front and settles the difference against actual usage afterwards; failed requests are not charged.
Images (nexith-image)#
| Item | Price |
|---|---|
| Generation | $0.10 / image |
| Editing | $0.10 / image |
- Reserved on submission, settled on completion, fully refunded on failure/cancellation.
- Artifact presigned URLs are valid for 7 days at no extra charge.
Account and charging#
| Topic | Description |
|---|---|
| Balance unit | USD, integer microdollars internally; Redis is the real-time source of truth and the DB is a periodic snapshot |
| Charge timing | Hard reservation before the request → settlement on completion; insufficient balance returns 402 before reaching the upstream |
| Failure refunds | Upstream failures, job failures, and request cancellations all automatically return the reservation |
| Plan multipliers | Free/Pro/Enterprise can have configurable multipliers; default 1.0, the console invoice is authoritative |
| Invoicing | Top-ups are invoiced by Waffo (Merchant of Record); API usage bills are available in the console |
Differences from OpenAI pricing#
- Nexith text prices are fixed as shown above, with no batch / flex / priority tiers.
- Caching applies only to requests whose upstream supports prefix caching; cache misses are billed at the full input price.
- Images are priced per image (currently fixed at 1024×1024), not per output token.