Chat from $0.30 per million tokens.
Chat and completions on fast open models, behind the API you already build on. Same key and balance as Voice.
pip install openai−from openai import OpenAIclient = OpenAI(api_key=os.environ["MSERVE_API_KEY"],
)stream = client.chat.completions.create(model="qwen3.8-27b", messages=messages, stream=True,)for chunk in stream:print(chunk.choices[0].delta.content or "", end="")
Your code already works.
Chat completions, completions, and responses through the official OpenAI SDKs. Streaming included.
A short list, on purpose.
Qwen3.8 27B for chat and Qwen3 Embedding 0.6B for search at launch. Every model we add is one we'd pick ourselves.
Nothing to reconcile at month end.
Your agent's speech, transcription, and reasoning draw from one balance, so one invoice covers the whole thing.
Pricing
| Model | Best for | Price |
|---|---|---|
Qwen3.8 27B | Chat | Input$0.30/ 1M tokens Output$2.50/ 1M tokens |
Qwen3 Embedding 0.6B | Search, retrieval | $0.01/ 1M tokens |
Licensed Apache-2.0. Commercial use included.
FAQ
Which endpoints are supported?
/v1/chat/completions, /v1/completions, /v1/responses, /v1/embeddings, and /v1/models. Streaming and tool calls included.
How is Text billed?
Per token. Input and output priced separately. Same balance as Voice and Batch.
Can I turn thinking off?
Yes. Pass reasoning_effort, or chat_template_kwargs: {"enable_thinking": false}. Thinking never leaks into content.
Start building for less.
One email when your key is ready. Nothing else.