Pricing
Pay per token from prepaid credits. No subscriptions, no seat fees. The table below is
rendered from the same catalog the API serves at
GET /models — what you see here is what you are billed.
Model pricing
All rates in USD per million tokens. Cached prompt tokens are billed at the discounted cached rate where the model supports prompt caching.
| Model | Prompt / Mtok | Completion / Mtok | Cached prompt / Mtok | Context | Max output | Features | Datacenters |
|---|---|---|---|---|---|---|---|
DeepSeek V4 Flashdeepseek/deepseek-v4-flash | $0.15 | $0.3 | $0.08 | 1.049M | 32.768K | tools, json_mode, reasoning, streaming, prompt_caching | eu-west, us-east, uae |
DeepSeek V4 Prodeepseek/deepseek-v4-pro | $1.83 | $3.66 | $0.35 | 1.049M | 32.768K | tools, json_mode, reasoning, streaming, prompt_caching | eu-west, us-east, uae |
DeepSeek V3.2deepseek/deepseek-v3.2 | $0.59 | $1.77 | $0.27 | 163.84K | 65.536K | tools, json_mode, reasoning, streaming, prompt_caching | eu-west, us-east, uae |
Qwen3.5 397B A17Bqwen/qwen3.5-397b-a17b | $0.79 | $4.73 | $0.58 | 262.144K | 65.536K | tools, vision, json_mode, reasoning, streaming, prompt_caching | eu-west, us-east, uae |
Qwen3 Coder 480B A35Bqwen/qwen3-coder | $1.05 | $5.12 | $1.05 | 1.049M | 65.536K | tools, json_mode, streaming, prompt_caching | eu-west, us-east, uae |
GLM 5.2z-ai/glm-5.2 | $1.47 | $4.62 | $0.28 | 1.049M | 131.072K | tools, json_mode, reasoning, streaming, prompt_caching | eu-west, us-east, uae |
GLM 5.1z-ai/glm-5.1 | $1.62 | $5.09 | $0.31 | 202.752K | 131.072K | tools, json_mode, reasoning, streaming, prompt_caching | eu-west, us-east, uae |
Gemma 4 31Bgoogle/gemma-4-31b-it | $0.41 | $1.02 | — | 262.144K | 16.384K | tools, vision, json_mode, reasoning, streaming | eu-west, us-east, uae |
Gemma 4 26B A4Bgoogle/gemma-4-26b-a4b-it | $0.16 | $0.63 | — | 262.144K | 16.384K | tools, vision, json_mode, reasoning, streaming | eu-west, us-east, uae |
Llama 4 Maverickmeta-llama/llama-4-maverick | $0.37 | $1.21 | $0.18 | 1.049M | 16.384K | tools, vision, json_mode, streaming | eu-west, us-east, uae |
Llama 3.3 70B Instructmeta-llama/llama-3.3-70b-instruct | $1.1 | $1.1 | $0.31 | 131.072K | 16.384K | tools, json_mode, streaming | eu-west, us-east, uae |
Llama 4 Scoutmeta-llama/llama-4-scout | $0.27 | $0.74 | $0.06 | 1.311M | 16.384K | tools, vision, json_mode, streaming | eu-west, us-east, uae |
Mistral Small 3.2 24Bmistralai/mistral-small-3.2-24b-instruct | $0.11 | $0.32 | $0.06 | 131.072K | 32.768K | tools, vision, json_mode, streaming, prompt_caching | eu-west, us-east, uae |
Mistral Large 3mistralai/mistral-large-2512 | $0.53 | $1.58 | $0.06 | 262.144K | 32.768K | tools, vision, json_mode, streaming, prompt_caching | eu-west, us-east, uae |
Kimi K2.6moonshotai/kimi-k2.6 | $1.26 | $4.83 | $0.39 | 262.144K | 32.768K | tools, vision, json_mode, reasoning, streaming, prompt_caching | eu-west, us-east, uae |
Kimi K2.7 Codemoonshotai/kimi-k2.7-code | $2 | $8.4 | $0.4 | 262.144K | 32.768K | tools, vision, json_mode, reasoning, streaming, prompt_caching | eu-west, us-east, uae |
Credits & top-up fees
Credits are workspace-scoped, denominated in USD, and never expire. Top-ups carry a service fee, shown transparently before you confirm:
| Payment method | Service fee | Example: $1,000 in credits | Minimum top-up |
|---|---|---|---|
| Card | 5.5% | $55.00 fee → $1055.00 charged | $5.00 |
| Crypto (USDC) | 5% | $50.00 fee → $1050.00 charged | $5.00 |
Buying $1,000 in credits by card charges $1,055.00; by crypto (USDC) $1,050.00 — you receive the full $1,000 in credits either way. Depending on your billing country and status, tax (e.g. UAE VAT, EU consumer VAT) may apply on top and is itemized at checkout.
Server tools
Server-tool executions (web search, web fetch, PDF parsing) are billed per use against workspace credits as separate, itemized ledger lines tied to the request that triggered them. Per-tool rates are published here before general availability of each tool.
Credits calculator
Estimate in both directions — tokens to dollars, and what a budget buys:
Estimated cost: $0.021 in credits
$10 buys ≈ 66.67M prompt tokens or ≈ 33.33M completion tokens on DeepSeek V4 Flash.
Estimates use uncached prompt rates from the catalog above. Actual billing is exact per-token math on the measured usage of each request — see Usage & Credits.