Merge branch 'huggingface:main' into main

Author: kefranabg

.github/workflows/api_inference_generate_documentation.yml

@@ -5,7 +5,6 @@ on:
   schedule:
     - cron: "0 3 * * *" # Every day at 3am
 
-
 concurrency:
   group: api_inference_generate_documentation
   cancel-in-progress: true
@@ -45,6 +44,7 @@ jobs:
           else
             echo "changes_detected=false" >> $GITHUB_ENV
           fi
+          rm changed_files.txt
 
       # Skip PR if only certain files are updated
       - name: Skip PR creation if no meaningful changes
@@ -68,10 +68,10 @@ jobs:
             pnpm update @huggingface/tasks@latest
             pnpm run generate
             ```
-            
+
             This PR was automatically created by the [Update API Inference Documentation workflow](https://github.com/huggingface/hub-docs/blob/main/.github/workflows/api_inference_generate_documentation.yml).
-            
+
             Please review the changes before merging.
           reviewers: |
             Wauplin
-            hanouticelina
+            hanouticelina

docs/api-inference/_redirects.yml

@@ -3,3 +3,4 @@ detailed_parameters: parameters
 parallelism: getting_started
 usage: getting_started
 faq: index
+rate-limits: pricing

docs/api-inference/_toctree.yml

@@ -5,8 +5,8 @@
     title: Getting Started
   - local: supported-models
     title: Supported Models
-  - local: rate-limits
-    title: Rate Limits
+  - local: pricing
+    title: Pricing and Rate limits
   - local: security
     title: Security
   title: Getting Started

docs/api-inference/getting-started.md

@@ -1,6 +1,6 @@
 # Getting Started
 
-The Serverless Inference API allows you to easily do inference on a wide range of models and tasks. You can do requests with your favorite tools (Python, cURL, etc). We also provide a Python SDK (`huggingface_hub`) to make it even easier.
+The Serverless Inference API allows you to easily do inference on a wide range of models and tasks. You can do requests with your favorite tools (Python, cURL, etc). We also provide a Python SDK (`huggingface_hub`) and JavaScript SDK (`huggingface.js`) to make it even easier.
 
 We'll do a minimal example using a [sentiment classification model](https://huggingface.co/cardiffnlp/twitter-roberta-base-sentiment-latest). Please visit task-specific parameters and further documentation in our [API Reference](./parameters).
 

docs/api-inference/index.md

@@ -8,14 +8,14 @@ Explore the most popular models for text, image, speech, and more — all with a
 
 ## Why use the Inference API?
 
-The Serverless Inference API offers a fast and free way to explore thousands of models for a variety of tasks. Whether you're prototyping a new application or experimenting with ML capabilities, this API gives you instant access to high-performing models across multiple domains:
+The Serverless Inference API offers a fast and simple way to explore thousands of models for a variety of tasks. Whether you're prototyping a new application or experimenting with ML capabilities, this API gives you instant access to high-performing models across multiple domains:
 
 * **Text Generation:** Including large language models and tool-calling prompts, generate and experiment with high-quality responses.
 * **Image Generation:** Easily create customized images, including LoRAs for your own styles.
 * **Document Embeddings:** Build search and retrieval systems with SOTA embeddings.
 * **Classical AI Tasks:** Ready-to-use models for text classification, image classification, speech recognition, and more.
 
-⚡ **Fast and Free to Get Started**: The Inference API is free with higher rate limits for PRO users. For production needs, explore [Inference Endpoints](https://ui.endpoints.huggingface.co/) for dedicated resources, autoscaling, advanced security features, and more.
+⚡ **Fast and Free to Get Started**: The Inference API is free to try out and comes with additional included credits for PRO users. For production needs, explore [Inference Endpoints](https://ui.endpoints.huggingface.co/) for dedicated resources, autoscaling, advanced security features, and more.
 
 ---
 

docs/api-inference/pricing.md

@@ -0,0 +1,20 @@
+# Pricing and Rate limits
+
+As a HF user, you get monthly credits to run the HF Inference API. The amount of credits you get depends on your type of account (Free or PRO or Enterprise Hub), see table below.
+You get charged for every inference request, based on the compute time x price of the underlying hardware.
+
+For instance, a request to [black-forest-labs/FLUX.1-dev](https://huggingface.co/black-forest-labs/FLUX.1-dev) that takes 10 seconds to complete on a GPU machine that costs $0.00012 per second to run, will be billed $0.0012.
+
+When your monthly included credits are depleted:
+- if you're a Free user, you won't be able to query the Inference API anymore,
+- if you're a PRO or Enterprise Hub user, you will get charged for the requests on top of your subscription. You can monitor your spending on your billing page.
+
+Note that HF Inference API is not meant to be used for heavy production applications. If you need to handle large numbers of requests, consider [Inference Endpoints](https://huggingface.co/docs/inference-endpoints) to have dedicated resources or [Inference Providers](https://huggingface.co/blog/inference-providers) for serverless usage.
+
+You need to be authenticated (passing a token or through your browser) to use the Inference API.
+
+
+| User Tier                 | Included monthly credits           |
+|---------------------------|------------------------------------|
+| Free Users                | subject to change, less than $0.10 |
+| PRO and Enterprise Users  | $2.00                              |

docs/api-inference/rate-limits.md

@@ -10,7 +10,7 @@ You can find:
 
 ## What do I get with a PRO subscription?
 
-In addition to thousands of public models available in the Hub, PRO and Enterprise users get higher [rate limits](./rate-limits) and free access to the following models:
+In addition to thousands of public models available in the Hub, PRO and Enterprise users get higher [included credits](./pricing) and access to the following models:
 
 <!-- Manually maintained hard-coded list based on https://github.com/huggingface-internal/api-inference/blob/main/master-rs/custom_config.yml -->
 
@@ -27,4 +27,4 @@ This list is not exhaustive and might be updated in the future.
 
 ## Running Private Models
 
-The free Serverless API is designed to run popular public models. If you have a private model, you can use [Inference Endpoints](https://huggingface.co/docs/inference-endpoints) to deploy it.
+The Serverless API is designed to run popular public models. If you have a private model, you can use [Inference Endpoints](https://huggingface.co/docs/inference-endpoints) to deploy it.

docs/api-inference/supported-models.md

@@ -30,7 +30,6 @@ For more details about the `automatic-speech-recognition` task, check out its [d
 ### Recommended models
 
 - [openai/whisper-large-v3](https://huggingface.co/openai/whisper-large-v3): A powerful ASR model by OpenAI.
-- [nvidia/canary-1b](https://huggingface.co/nvidia/canary-1b): A powerful multilingual ASR and Speech Translation model by Nvidia.
 - [pyannote/speaker-diarization-3.1](https://huggingface.co/pyannote/speaker-diarization-3.1): Powerful speaker diarization model.
 
 Explore all available models and find the one that suits you best [here](https://huggingface.co/models?inference=warm&pipeline_tag=automatic-speech-recognition&sort=trending).

docs/api-inference/tasks/automatic-speech-recognition.md

@@ -23,13 +23,15 @@ This is a subtask of [`text-generation`](https://huggingface.co/docs/api-inferen
 
 - [google/gemma-2-2b-it](https://huggingface.co/google/gemma-2-2b-it): A text-generation model trained to follow instructions.
 - [meta-llama/Meta-Llama-3.1-8B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3.1-8B-Instruct): Very powerful text generation model trained to follow instructions.
-- [microsoft/Phi-3-mini-4k-instruct](https://huggingface.co/microsoft/Phi-3-mini-4k-instruct): Small yet powerful text generation model.
+- [microsoft/phi-4](https://huggingface.co/microsoft/phi-4): Powerful text generation model by Microsoft.
+- [PowerInfer/SmallThinker-3B-Preview](https://huggingface.co/PowerInfer/SmallThinker-3B-Preview): A very powerful model with reasoning capabilities.
 - [Qwen/Qwen2.5-7B-Instruct](https://huggingface.co/Qwen/Qwen2.5-7B-Instruct): Strong text generation model to follow instructions.
+- [Qwen/Qwen2.5-Coder-32B-Instruct](https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct): Text generation model used to write code.
 
 #### Conversational Vision-Language Models (VLMs)
 
-- [meta-llama/Llama-3.2-11B-Vision-Instruct](https://huggingface.co/meta-llama/Llama-3.2-11B-Vision-Instruct): Powerful vision language model with great visual understanding and reasoning capabilities.
 - [Qwen/Qwen2-VL-7B-Instruct](https://huggingface.co/Qwen/Qwen2-VL-7B-Instruct): Strong image-text-to-text model.
+- [Qwen/QVQ-72B-Preview](https://huggingface.co/Qwen/QVQ-72B-Preview): Image-text-to-text model with reasoning capabilities.
 
 ### API Playground
 
@@ -208,11 +210,11 @@ To use the JavaScript client, see `huggingface.js`'s [package reference](https:/
 
 <curl>
 ```bash
-curl 'https://api-inference.huggingface.co/models/meta-llama/Llama-3.2-11B-Vision-Instruct/v1/chat/completions' \
+curl 'https://api-inference.huggingface.co/models/Qwen/Qwen2-VL-7B-Instruct/v1/chat/completions' \
 -H 'Authorization: Bearer hf_***' \
 -H 'Content-Type: application/json' \
 --data '{
-    "model": "meta-llama/Llama-3.2-11B-Vision-Instruct",
+    "model": "Qwen/Qwen2-VL-7B-Instruct",
     "messages": [
 		{
 			"role": "user",
@@ -262,7 +264,7 @@ messages = [
 ]
 
 stream = client.chat.completions.create(
-    model="meta-llama/Llama-3.2-11B-Vision-Instruct", 
+    model="Qwen/Qwen2-VL-7B-Instruct", 
 	messages=messages, 
 	max_tokens=500,
 	stream=True
@@ -300,7 +302,7 @@ messages = [
 ]
 
 stream = client.chat.completions.create(
-    model="meta-llama/Llama-3.2-11B-Vision-Instruct", 
+    model="Qwen/Qwen2-VL-7B-Instruct", 
 	messages=messages, 
 	max_tokens=500,
 	stream=True
@@ -323,7 +325,7 @@ const client = new HfInference("hf_***");
 let out = "";
 
 const stream = client.chatCompletionStream({
-	model: "meta-llama/Llama-3.2-11B-Vision-Instruct",
+	model: "Qwen/Qwen2-VL-7B-Instruct",
 	messages: [
 		{
 			role: "user",
@@ -365,7 +367,7 @@ const client = new OpenAI({
 let out = "";
 
 const stream = await client.chat.completions.create({
-	model: "meta-llama/Llama-3.2-11B-Vision-Instruct",
+	model: "Qwen/Qwen2-VL-7B-Instruct",
 	messages: [
 		{
 			role: "user",