Add Maison Token provider and Qwen3.8-27B mapping

[kartik-sk]

Summary

• add Maison Token as an OpenAI-compatible US inference provider
• add the canonical Qwen3.8-27B model definition
• add Maison Token's Qwen3.8-27B mapping with token pricing, rate limits, and availability

Pricing

• input: $0.20 per 1M tokens
• cache read: $0.05 per 1M tokens
• output: $1.35 per 1M tokens

Validation

• .venv/bin/python validate_registry.py
• Validation passed for all 403 registry files

[madeburo]

Hi @kartik-sk
Thanks for the contribution. The Qwen3.8-27B model entry is generally correct and local registry validation passes.
Before merging, please address the following:

1. Please provide public documentation for MaisonToken’s direct API access, API key signup, pricing, rate limits, and regional availability. The current website only contains legal pages.

2. The MaisonToken Terms state that the model is provided through OpenRouter and that pricing is displayed and billed there. The public OpenAPI schema also identifies the service as MaisonToken OpenRouter Provider. Please clarify whether https://api.maisontoken.co/v1 is a publicly available standalone API or only an internal OpenRouter origin.

3. MaisonToken is not currently listed among the public OpenRouter endpoints for Qwen3.8-27B. Therefore, the submitted $0.20 / $1.35 pricing, 384 RPM / 440K TPM limits, and us-west-2 availability cannot currently be verified.

4. The canonical model supports 262,144 tokens natively and up to 1,000,000 tokens with YaRN. Please set the model’s context_window to 1000000. If the MaisonToken deployment is limited to 262,144 tokens, set context_window_override: 262144 in the mapping.

5. Please use the official brand spelling MaisonToken rather than Maison Token.

If MaisonToken is only accessible through OpenRouter, please remove the standalone provider and mapping entries. The canonical model entry can remain, and an OpenRouter mapping may be considered separately using publicly verifiable deployment data.
Once these points are addressed, we’ll be happy to review the PR again. Thank you!