Subject: Highreso (Japan) — request to be listed as an Inference Provider
Hi Hugging Face Inference Providers team,
I’m writing on behalf of Highreso Co., Ltd., a GPU cloud operator based in Japan (data centers in Kagawa and Saga). We would like to be listed as an Inference Provider on the Hub.
What we offer: an OpenAI-compatible chat completions API at https://api.highreso.ai/hf/v1 (streaming, tool calling, structured output), served on Highreso-owned NVIDIA B200 clusters located in Japan. All data stays in Japan. Initial models: zai-org/GLM-5.3 (FP8) and Qwen/Qwen3-8B, with moonshotai/Kimi-K3 to follow. Measured single-user speed on our B200 nodes: GLM-5.3 108 to 149 tokens/s (p50) at 4 to 8 concurrent users. We list at the standard provider prices shown on the Hub. We expose a per-request Inference-Id header, a billing endpoint (POST /hf/v1/billing) that returns cost in nano-USD per request ID, and pricing plus context_length on /hf/v1/models.
Status: PRs for huggingface.js (provider “highreso”) and huggingface_hub are open:
Hub organization: highreso-jp (Team plan). Company site: highreso.jp. Contact: inference@highreso.co.jp
Could you enable our organization server-side so we can register model mappings via /api/partners/highreso/models? Happy to provide a test key for your validation runs.
Thank you,
Highreso Co., Ltd.