DeepInfra raises $107M Series B to scale the inference cloud — read the announcement

DeepInfra is officially live as an Inference Provider on the Hugging Face Hub. You can now call DeepInfra-hosted models directly from Hugging Face model pages, through our OpenAI-compatible router (use it with any OpenAI SDK), or via the Hugging Face SDKs in Python and JavaScript.
Hugging Face's Inference Providers system lets developers run inference against partner platforms without leaving the Hub. As of today, DeepInfra is one of those partners.
At launch, we support chat completion and text generation tasks. That covers most open-weight LLMs people deploy in production — DeepSeek V4, Kimi-K2.6, GLM-5.1, Llama, Qwen, Mistral, and many more. Support for our other model categories (text-to-image, text-to-video, embeddings, speech) will roll out next.
You can browse every DeepInfra-supported model here: 👉 huggingface.co/models?inference_provider=deepinfra
You have two ways to authenticate, and both work with the same code.
Option 1 — Use your DeepInfra API key. Add it to your Hugging Face provider settings. Requests go directly to DeepInfra and are billed to your DeepInfra account at standard rates.
Option 2 — Use your Hugging Face token. Hugging Face will route your request to DeepInfra and bill it to your HF account. PRO users get $2 of inference credits each month; free users get a small monthly quota.
from huggingface_hub import InferenceClient
client = InferenceClient()
completion = client.chat.completions.create(
model="deepseek-ai/DeepSeek-V4-Pro:deepinfra",
messages=[
{"role": "user", "content": "Write a Fibonacci function with memoization."}
],
)
print(completion.choices[0].message)
import { InferenceClient } from "@huggingface/inference";
const client = new InferenceClient(process.env.HF_TOKEN);
const completion = await client.chatCompletion({
model: "deepseek-ai/DeepSeek-V4-Pro:deepinfra",
messages: [{ role: "user", content: "Hello!" }],
});
console.log(completion.choices[0].message);
The Hugging Face router is OpenAI-compatible, so existing OpenAI code works with one line changed — point base_url at the HF router:
from openai import OpenAI
client = OpenAI(
base_url="https://router.huggingface.co/v1",
api_key=os.environ["HF_TOKEN"],
)
completion = client.chat.completions.create(
model="deepseek-ai/DeepSeek-V4-Pro:deepinfra",
messages=[{"role": "user", "content": "Hello!"}],
)
The only thing that changes is the :deepinfra suffix on the model id.
If you already use DeepInfra, nothing changes — your existing API and account work exactly as they always have. What's new is reach.
Muse Glimmer 30B Benchmarks: Speed, Latency & Cost per Task<p>Updated November 2026. Muse Glimmer 30B is Meta’s open-weight agent model, built to run on a single GPU and distilled from Meta’s proprietary Muse Spark. This page collects the independent numbers for it: Artificial Analysis figures for the model and for its DeepInfra endpoint, compared with the medians Artificial Analysis publishes for open-weight models of […]</p>
Best API Providers for Wan 3.0 Integration<p>Wan3.0 is Alibaba’s all-in-one video generation model. It accepts text, images, video, audio, documents, and web pages, and it produces up to 30 seconds of 1080P video with synchronized audio. Production use raises practical questions. Video jobs run asynchronously, reference inputs can be large, and each provider exposes a different slice of the model’s features […]</p>
GLM-5.3 Model Overview & Integration Guide<p>GLM-5.3 is a large-scale reasoning model from Z.ai, built for complex software engineering and long-horizon agent tasks. It handles text input and output across a 1,048,576-token context window, and is one of several hundred models in DeepInfra’s catalog. The model uses the same base as GLM-5.2. Every gain comes from post-training, which is where Z.ai […]</p>
© 2026 DeepInfra. All rights reserved.