Inference API for open-weight models

We serve open-weight language models through an OpenAI-compatible API priced per million tokens. We work with routers, marketplaces and direct customers.

Models and pricing

USD per 1M tokens. Usage is billed on the token counts returned in each response's usage object.

API

• OpenAI-compatible /v1/chat/completions and /v1/models
• Streaming (SSE), tool and function calling
• Prompt caching: cached input tokens are reported in usage.prompt_tokens_details.cached_tokens and billed at the cached rate
• Per-key rate limits, with a Retry-After header on 429 responses

curl https://<your-netra-endpoint>/v1/chat/completions \
  -H "Authorization: Bearer $NETRA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"qwen3.8-27b","messages":[{"role":"user","content":"Hello"}]}'

We issue endpoints and keys during onboarding.

Infrastructure

We serve on NVIDIA and AMD hardware with our in-house serving engine.

Data policy

• We do not train on customer prompts or completions.
• We do not store prompts or completions beyond short-lived prompt caching.
• We keep request metadata (timestamp, model, token counts, latency, status) for billing, abuse prevention and reliability.

For routers and marketplaces

We provide dedicated keys, usage reporting per key and an operational contact. To list Netra as a provider, email us.

Contact

[email protected]