Pricing
Transparent pricing
Pay only for what you use, at published per-unit prices that are the same on every plan.
Inference pricing
Chat models, by author. Prices are per 1 million tokens.
DeepSeek
Efficient open-weight models for reasoning, coding and long-context chat.
| Model | Context | Input | Cached input | Output |
|---|---|---|---|---|
| DeepSeek V4.1 Flash | 1049K | $0.30 | $0.006 | $1.40 |
| deepseek-v4-flash-0731 | 1049K | $0.14 | $0.00 | $0.28 |
Gemma, Google’s open-weight family, with strong multilingual performance.
| Model | Context | Input | Cached input | Output |
|---|---|---|---|---|
| gemma-4-31b-it | 262K | $0.27 | $0.00 | $0.76 |
Infiniatechnologies
Open-weight models served in your region.
| Model | Context | Input | Cached input | Output |
|---|---|---|---|---|
| nia-1.0 | 131K | $1.40 | $0.26 | $4.40 |
| saqr-27b | 262K | $0.27 | $0.00 | $0.76 |
Microsoft
Open-weight models served in your region.
| Model | Context | Input | Cached input | Output |
|---|---|---|---|---|
| fara1.5-27b | 262K | $0.27 | $0.00 | $0.76 |
OpenAI
GPT-OSS, OpenAI’s Apache-licensed open-weight models.
| Model | Context | Input | Cached input | Output |
|---|---|---|---|---|
| GPT-OSS 120B | — | $0.15 | $0.00 | $0.60 |
| GPT-OSS 20B | — | $0.05 | $0.00 | $0.20 |
Qwen
Alibaba’s Qwen family, from compact chat models to multimodal models with tool use.
| Model | Context | Input | Cached input | Output |
|---|---|---|---|---|
| qwen3-omni-30b-a3b-instruct | 66K | $0.43 | $0.00 | $1.66 |
| qwen3.8-27b | 262K | $0.42 | $0.00 | $3.00 |
Tencent
Open-weight models served in your region.
| Model | Context | Input | Cached input | Output |
|---|---|---|---|---|
| hunyuanocr | 16K | $0.14 | $0.00 | $0.57 |
Zai Org
Open-weight models served in your region.
| Model | Context | Input | Cached input | Output |
|---|---|---|---|---|
| glm-ocr | 131K | $0.04 | $0.00 | $0.08 |
Embeddings, speech and image
Each model type is priced in its own unit.
Embeddings
Vector embeddings for search and retrieval. Priced per 1M input tokens.
| qwen3-embedding-8b | $0.10 per 1M tokens |
Speech to text
Transcription, and translation into English. Priced per minute of audio.
| qwen3-asr-1.7b | $0.006 per minute |
| sravaani-1.0 | $0.006 per minute |
Text to speech
Speech synthesis from text. Priced per 1M characters.
| chatterbox-multilingual-packs | $15.00 per 1M chars |
| qwen3.5-tts-emirati | $15.00 per 1M chars |
Image
Image generation. Priced per image.
| sd3.5-large | $0.04 per image |
Video
Video generation. Priced per second.
| ltx-2.5 | $0.05 per second |
Provisioned throughput
For steady, business-critical workloads, reserve capacity on the models you depend on so speed and volume stay predictable. Priced by agreement for the capacity you reserve.
Frequently asked questions
How billing, plans and capacity work.
How does billing work?
You pay for what you use. Chat models are charged per token for input and output; other models are charged in their own unit, such as per minute of audio or per image. Individual and Enterprise accounts pay from prepaid credit; Sovereign customers are invoiced monthly under contract.
What is cached input?
The rate for prompt tokens the model has recently seen, such as a repeated system prompt. It replaces the input rate for those tokens, and writing to the cache is not charged.
Does my plan change the price?
No. Individual, Enterprise and Sovereign plans change the service level you get, not the price of usage. Every plan pays the same published rate.
Can I set spending limits?
Yes. Admins can set spending caps on each project in the console, so a key cannot spend beyond what you allow.
Do prices differ between regions?
A model’s price can differ by region. The console shows the price for your project’s region before you make a call.
How is provisioned throughput priced?
By agreement, for the capacity you reserve on a model. Talk to our team about your volume and the models you need.
Service tiers
Tiers set the service level and support you get. The price of usage is the same on every tier.
Individual
For developers and researchers working on their own.
- Every model at the published price
- Pay as you go with prepaid credit
- Sign up and make your first call in minutes
- Documentation and community support
Enterprise
For companies running AI in production.
- Everything in Individual
- Organisation with projects, roles and spending caps
- Uptime commitment with service credits
- Single sign-on and user provisioning
- Provisioned throughput for steady, high-volume workloads
- Priority support with response targets
Sovereign
For government departments, ministries and public authorities.
- Everything in Enterprise
- Named account manager
- Full due diligence completed with our team
- Signed SLA, contract terms and monthly invoicing
Plans set the service level, not the price of usage. Every plan pays the same per-unit price for the same model.
Planning a large volume?
Talk to us about provisioned throughput and contract terms.