Spaces:

---

huggingface

/

HuggingDiscussions

Like 105

Running

Love that I can call DeepSeek R1 directly from the Hub 🔥

from huggingface_hub import InferenceClient

client = InferenceClient(
    provider="together",
    api_key="xxxxxxxxxxxxxxxxxxxxxxxx"
)

messages = [
    {
        "role": "user",
        "content": "What is the capital of France?"
    }
]

completion = client.chat.completions.create(
    model="deepseek-ai/DeepSeek-R1", 
    messages=messages, 
    max_tokens=500
)

print(completion.choices[0].message)

Is it possible to set a monthly payment budget or rate limits for all the external providers? I don't see such options in billings tab. In case a key is or session token is stolen, it can be quite dangerous to my thin wallet:(

@benhaotang you already get spending notifications when crossing important thresholds ($10, $100, $1,000) but we'll add spending limits in the future

@benhaotang you already get spending notifications when crossing important thresholds ($10, $100, $1,000) but we'll add spending limits in the future

Thanks for your quick reply, good to know!

Would be great if you could add Nebius AI Studio to the list :) New inference provider on the market, with the absolute cheapest prices and the highest rate limits...

Could be good to add featherless.ai

TitanML !!

Hi Hugging Face team,

We're IronStratum : we run a per-token, OpenAI-compatible inference API at https://api.ironstratum.com/v1.

We have several models exclusive to our platform, on GPU capacity we operate ourselves, alongside a catalog covering the DeepSeek, Qwen, GLM, and MiniMax families, with fresh builds listed fast (we listed
DeepSeek V4.1 Flash 1017 two days after its release) and specialty lanes for OCR, classification, voice, robotics and image generation. Several of these are Hub models with no inference provider today. They ship on dedicated action and vision routes with per-call and per-image metering.

We'd like to be listed as an Inference Provider. Following the registration guide, the pieces on our side are in place:

• Chat completions follow the OpenAI API shape
• /v1/models is anonymously readable and exposes pricing (USD per million input/output tokens) and context_length, in the format the guide expects.
• The billing endpoint is implemented to the documented batch protocol:
  POST with requestIds, responses in nano-USD integers, idempotent, inside the 30-minute window, authenticated by a dedicated key. The response header carrying our request id is x-request-id
• Tool calling and structured output (response_format with json_schema) are both live through our API and verified on our serving lanes.

Our organization on the Hub is iron-stratum (Team plan). We're ready to open the huggingface.js PR, register our initial mappings in staging, and follow with the huggingface_hub PR and a provider page for the docs.

What's the best way to start the huggingface.js PR first, or the server-side enablement with you first? Happy to share details on infrastructure, latency, or pricing.

Thanks

Hi Hugging Face team,

We are Apex Fusion (https://huggingface.co/apexfusion) and would like to be
registered as an Inference Provider. Conversational task only for now,
OpenAI-compatible API, served from our own GPU server in Helsinki, Finland.

• huggingface.js PR: https://github.com/huggingface/huggingface.js/pull/2686
  (check, lint and format pass; @Wauplin and @SBrandeis are tagged)
• Endpoint: https://hf-edge.vector.apexfusion.org (/v1/chat/completions,
  streaming, tool calling, structured output; /v1/models exposes pricing
  and context_length)
• Inference-Id response header and the billing endpoint are implemented
• First model: Qwen/Qwen3.8-27B

We will send the billing URL and the bearer token through a private channel
once you tell us where. Happy to adjust anything in the PR.

Thanks,
Apex Fusion

Negbeta (EU inference provider): request to join Inference Providers with DeepSeek V4.1 Flash

Hi Hugging Face team,

We'd like to join Inference Providers with DeepSeek V4.1 Flash 1017, including an EU-hosted option for Hub users who need EU processing.

Negbeta is a new inference provider: top-quality inference with US and EU capacity, zero data retention and strict EU compliance, at a competitive price.

• Full 1M context on DeepSeek, fast time to first token, strong prompt caching for agentic workloads, reliable under load; OpenAI-compatible, streaming, tool calls, structured outputs, reasoning, X-Request-Id on every response.
• We'll do the integration work: client PRs, model mappings, billing endpoint.

Qwen3.8 27B is also in our catalogue, and we're ramping up further models to serve the demand. We're onboarded with one router, with others in progress.

What would you like from us to get started?

Endpoint: https://inference.negbeta.com/v1/models · Docs: https://inference.negbeta.com/docs

Thomas Ip
Founder, Negbeta Limited
info@negbeta.com · https://negbeta.com

Prima Era: request to join Inference Providers (GLM-5.3, Kimi-K3)

Hi Hugging Face team,

We're Prima Era (https://primaera.xyz, org https://huggingface.co/primaera on the Team plan) and would like to be registered as an Inference Provider for the conversational task, with an OpenAI-compatible API.

• Endpoint: https://api.primaera.xyz/v1
• Hub model: zai-org/GLM-5.3, our model id glm-5.3 (FP8, 1,048,576-token context, streaming, tool calls, reasoning_content)
• List price per million tokens: $0.55 input, $0.18 cached input, $1.05 output
• We can also serve moonshotai/Kimi-K3 on the same GPUs and would like to map it too.
• NVIDIA B300 GPUs we operate in the US; we don't log prompts or completions.

Already in place for your requirements:

• pricing (USD per million input and output tokens) and context_length in /v1/models
• a unique X-Request-Id header on every response, streaming included
• billing API: POST https://api.primaera.xyz/v1/request-costs with {"requestIds": [...]}, returning costNanoUsd per request, with the same bearer auth as inference

Next we'll open the huggingface.js PR. Who should we coordinate with? Contact: hf@primaera.xyz

Thanks!

Hi Hugging Face provider-onboarding team — following up on PyAI’s existing Hear/Speak request. Could you help us identify the current blocker and the next concrete qualification step?

Before we implement a native provider adapter, we need confirmation of API-only ASR/TTS eligibility, the required Hub model mappings, the ASR/TTS request and response contract, request-cost billing requirements, and any organization-plan prerequisites. Our working direct integrations are linked above; we understand they do not establish native Inference Providers approval or routing.

Is there a technical owner who can confirm the checklist and tell us what evidence or corrections are needed from PyAI? If audio-provider onboarding is currently unavailable or the request is simply queued, please let us know so we can plan appropriately. Thank you!

Runcrate: request to join Inference Providers when onboarding reopens

Hey Hugging Face team,

Jish here, founder of Runcrate (runcrate.ai). We run an OpenAI-compatible inference API for open-weight models, and we've built the provider side to your spec: the billing endpoint answers cost per request id in nano-USD, and our models list carries live pricing.

Tasks we'd bring: chat (LLM and VLM), feature extraction, text to image and speech to text.

We saw on huggingface.js#2698 that new-provider onboarding is paused for now. Could you keep us on the list for when it opens again? Happy to open the JS client PR whenever suits you.

Best,
Jish

Montagea — Request to join Hugging Face Inference Providers (FastH3 Omni)

Hi Hugging Face team, @Wauplin @SBrandeis ,

We are building Montagea, an AI video creation platform. Our API is deployed, and we'd like to join Inference Providers, initially with FastH3 Omni.

Website: https://montagea.com
API docs: https://montagea.com/docs/api
HF organization: https://huggingface.co/Montagea-com
Contact: tim@montagea.com

Initial model:
FastVideo/FastVideo-FastH3-Omni-8-Step-V1
Our API model ID: minimax-h3

Infrastructure:
We operate GPU instances on Google Cloud using NVIDIA RTX PRO 6000 Blackwell GPUs and UniServe.

Capabilities:

• Text-to-video and image-to-video
• Reference image, video and audio inputs
• Asynchronous video submission, polling and retrieval

Our implementation includes:

• Provider API: https://api.montagea.com/api/partners/huggingface/v1
• Inference-Id response headers
• Per-request billing lookup in integer nano-USD
• Local JavaScript/Python adapters; upstream SDK registration is pending

Could you please confirm:

1. Are you currently onboarding new video inference providers?
2. How should FastH3 Omni's reference inputs map to HF tasks and SDK parameters?
3. What timeout and validation rules apply to asynchronous video generation?
4. What are the current organization-plan, settlement and server-side registration requirements?

We can arrange a scoped evaluation key privately and prepare the required SDK PRs, model mappings and documentation.

Thank you!