Pay as you go
Serverless endpoints
Per-token access to frontier open source models. Call an endpoint and go: zero setup.
- Frontier Open Source Models
- No minimums
- Community support
01 / Plans
Pay per token
Frontier open-weight models on an OpenAI-compatible API. No per-seat fee.
Early access
Elastic Endpoints
Private, optimized endpoints priced per million tokens, with per-workload tuning.
- Private optimized endpoints
- Per-workload tuning
- Dedicated support
Reserved capacity
Dedicated deployments
Reserved GPUs sized to your roadmap, with negotiated latency SLAs.
- Reserved GPUs, sized to you
- Negotiated latency SLA
- Dedicated support
Lowest rate
Batch
High-throughput offline jobs at the lowest per-token rate, on spare fleet capacity.
- Lowest per-token rate
- Millions of requests per job
- Great for evals & embeddings
02 / Elastic Endpoints
Per-token model pricing
Pay only for what you use. No minimums, no rate limits.
Price per 1M tokens
| Model | Input | Output | Cache read |
|---|---|---|---|
| GLM-5.3 | $1.40 | $4.40 | $0.26 |
| Kimi K3 | $3.00 | $15.00 | $0.30 |
| DeepSeek-V4-Flash | $0.14 | $0.28 | $0.07 |
| Qwen3.6 35B-A3B | $0.15 | $1.00 | $0.05 |
03 / Questions
How does pay as you go work?
Serverless endpoints are billed per token with no minimums. Call an endpoint and go: zero setup.
What are Elastic Endpoints?
Private, optimized endpoints priced per million tokens, with per-workload tuning and dedicated support.
When do I need dedicated deployments?
Reserved GPUs sized to your roadmap, with negotiated latency SLAs and dedicated support.
What is Batch?
High-throughput offline jobs at the lowest per-token rate, on spare fleet capacity. Built for millions of requests per job, evals, and embeddings.
Are prompts or responses retained?
No. Every option uses strict zero data retention for API inputs and outputs, and Prism never uses them for training. See the Privacy Policy for details.
04 / Deploy