Hitchhikethe internet
Directory/LLM models/Hy3 Preview
LLM models · Model

Hy3 Preview

Researched

Hy3-preview is a 299B-parameter text generation model by Tencent, released as a preview on Hugging Face via DeepInfra. It supports tool calling, structured output, configurable reasoning effort, and a 262,144-token context window at $0.14/$0.58 per 1M input/output tokens.

Online Checked Follow updates
Official model pageView model card
Official site snapshots

See the official site at a glance

Read-only public captures of Hy3 Preview’s homepage and verified pricing page. Screenshots are dated, never live embeds, and open full-screen.

Visit live site
Homepage · captured Jul 23, 2026
1 earlier capture
Pricing page · captured Jul 23, 2026
At a glance

In one minute

Start here for the decision-making essentials: what Hy3 Preview does, who it is for, how it is accessed, and the first-party sources behind this profile.

Pricing2 options

$0.14 per 1M input tokens

$0.58 per 1M output tokens

Platforms
Hosted on Hugging FacedeepinfraHugging Face
API accessNot public
FoundedNot disclosed by source
AvailabilityWeb / remote
LicenseProprietary

Best suited to

Source-backed fit
Long-context text generation workloads Tool-calling agent and workflow applications Structured data extraction tasks Cost-sensitive reasoning-enabled inference Evaluations and benchmarking on preview models
Model registry

Specifications and best fit

Source backed

Published specifications and use cases for Hy3 Preview. Repository dates are kept separate from the model's original release date.

Developer
tencent
Family
Not stated
Original release
Not verified
Repository created
2026-04-13
Parameters
299B
Context window
262,144 tokens
Architecture
Not stated
License
other
Weights
Not stated
Modalities
Not stated
Languages
Not stated
Repository
tencent/Hy3-preview

Good fit for

  • Text Generation
Decision support

Common questions and adoption checks

6 sourced answers

Short answers to the questions buyers and builders commonly ask about Hy3 Preview. Each answer cites the shared ledger below, where every source is listed once.

01What does Hy3 Preview say it can do?

Text Generation

tencent/Hy3-preview Text Generation • 299B
Ledger citation[4] huggingface.co/models
02What use cases does Hy3 Preview describe?

Text Generation

tencent/Hy3 Text Generation • 299B • Updated 2 days ago • 14.1k • • 857
Ledger citation[5] huggingface.co/models
03What pricing information is available for Hy3 Preview?

$0.14 per 1M input tokens · $0.58 per 1M output tokens

tencent/Hy3 Hy3 deepinfra $0.14 $0.58 262,144 0.29 42 Yes Yes
Ledger citation[3] huggingface.co/inference/models
04Does Hy3 Preview document API access?

Text Generation Inference (TGI) is an optimized toolkit for serving language models · Text Embeddings Inference (TEI) is an optimized toolkit for serving embeddings models

Text Generation Inference Serve language models with TGI optimized toolkit
Ledger citation[1] huggingface.co/docs
05What integrations does Hy3 Preview document?

Transformers is a library providing state-of-the-art AI models for PyTorch

Transformers State-of-the-art AI models for PyTorch
Ledger citation[1] huggingface.co/docs
06How can Hy3 Preview be deployed or accessed?

Hugging Face offers Inference Endpoints (dedicated) for deploying models on fully managed · Models can be trained and deployed on AWS using Hugging Face Deep Learning Containers · Hugging Face models can be deployed on Microsoft Azure · Hugging Face models can be trained and deployed on Google Cloud

Inference Endpoints (dedicated) Deploy models on dedicated & fully managed infrastructure on HF
Ledger citation[1] huggingface.co/docs
Decision guide

Capabilities and operating fit

LLM models

This profile connects the jobs Hy3 Preview is described as handling with its delivery model, access options and the subjects used to match it to related products in this directory.

Common use cases

  • Text generation
  • Tool-use / function calling
  • Structured (JSON-style) output generation
  • Configurable reasoning tasks (high, low, no_think)
  • Long-context processing up to 262,144 tokens
  • Text Generation

Verified capabilities

Access signals

Pricing model
$0.14 per 1M input tokens · $0.58 per 1M output tokens
API
Not publicly listed
Source links
16 recorded
Source-backed

Verified facts

Updated July 22, 2026

Each fact points to a recorded source, making it easy to distinguish verified product information from claims that need checking.

Official website

HTTP 200 verified twice

[2]huggingface.co/tencent/Hy3-preview
First-party description

tencent/Hy3-preview · Hugging Face

[2]huggingface.co/tencent/Hy3-preview
Source-supported facts

Model: tencent/Hy3-preview · Company: Tencent · Model version: preview

[2]huggingface.co/tencent/Hy3-preview
Model

Hy3-preview

[2]huggingface.co/tencent/Hy3-preview
Company

Tencent

[2]huggingface.co/tencent/Hy3-preview
Model version

preview

[2]huggingface.co/tencent/Hy3-preview
Tool use

Supports tool calling via special tokens (<tool_calls>, <tool_call>, <tool_sep>, <arg_key>, <arg_value>, <tool_responses>, <tool_response>)

[2]huggingface.co/tencent/Hy3-preview
Reasoning

Supports a reasoning mode with configurable effort levels (high, low, no_think) and think tokens (<think>, , <｜reasoning_mode｜>)

[2]huggingface.co/tencent/Hy3-preview
View 28 more verified facts
Platform

Hosted on Hugging Face

[2]huggingface.co/tencent/Hy3-preview
Platform

Hugging Face Hub hosts 200,000+ models via 10+ Inference partners

[1]huggingface.co/docs
Deployment

Hugging Face offers Inference Endpoints (dedicated) for deploying models on fully managed infrastructure

[1]huggingface.co/docs
Deployment

Models can be trained and deployed on AWS using Hugging Face Deep Learning Containers

[1]huggingface.co/docs
Deployment

Hugging Face models can be deployed on Microsoft Azure

[1]huggingface.co/docs
Deployment

Hugging Face models can be trained and deployed on Google Cloud

[1]huggingface.co/docs
Api

Text Generation Inference (TGI) is an optimized toolkit for serving language models

[1]huggingface.co/docs
Api

Text Embeddings Inference (TEI) is an optimized toolkit for serving embeddings models

[1]huggingface.co/docs
Open source

Hugging Face commits to advancing and democratizing AI through open source and open science

[1]huggingface.co/docs
Support

Hugging Face offers Enterprise Support as a paid offering

[1]huggingface.co/docs
Fine tuning

PEFT provides parameter-efficient finetuning for large language models

[1]huggingface.co/docs
Integration

Transformers is a library providing state-of-the-art AI models for PyTorch

[1]huggingface.co/docs
Quantization

Bitsandbytes is offered to optimize and quantize models

[1]huggingface.co/docs
Model

tencent/Hy3

[3]huggingface.co/inference/models
Company

tencent

[3]huggingface.co/inference/models
Platform

deepinfra

[3]huggingface.co/inference/models
Pricing

$0.14 per 1M input tokens

[3]huggingface.co/inference/models
Pricing

$0.58 per 1M output tokens

[3]huggingface.co/inference/models
Context window

262,144 tokens

[3]huggingface.co/inference/models
Tool use

Supported (Yes)

[3]huggingface.co/inference/models
Structured output

Supported (Yes)

[3]huggingface.co/inference/models
Model

tencent/Hy3-preview

[4]huggingface.co/models
Company

Tencent

[4]huggingface.co/models
Capability

Text Generation

[4]huggingface.co/models
Platform

Hugging Face

[4]huggingface.co/models
Model

tencent/Hy3

[5]huggingface.co/models
Parameter count

299B

[5]huggingface.co/models
Use case

Text Generation

[5]huggingface.co/models
Practical capabilities

What it helps with

6 documented areas

A concise view of the jobs, capabilities and integrations described in the recorded product sources.

Use case

Text generation

Use case

Tool-use / function calling

Use case

Structured (JSON-style) output generation

Use case

Configurable reasoning tasks (high, low, no_think)

Use case

Long-context processing up to 262,144 tokens

Integration

Transformers is a library providing state-of-the-art AI models for PyTorch

Availability

Where it runs and where to get it

Source checked

Documented product formats, platforms and official distribution destinations. Availability can vary by region and plan.

Cost / license

$0.14 per 1M input tokens · $0.58 per 1M output tokensHugging Face commits to advancing and democratizing AI through open source and open scienc

Platforms

Hosted on Hugging FaceHugging Face Hub hosts 200,000+ models via 10+ Inference partnersdeepinfraHugging Face
Implementation details

Adoption notes

DeploymentHugging Face offers Inference Endpoints (dedicated) for deploying models on fully managed · Models can be trained and deployed on AWS using Hugging Face Deep Learning Containers · Hugging Face models can be deployed on Microsoft Azure · Hugging Face models can be trained and deployed on Google Cloud
LicenseHugging Face commits to advancing and democratizing AI through open source and open scienc
Model supportHy3-preview · tencent/Hy3 · tencent/Hy3-preview
Data controlNot disclosed by source
Learning curveIntermediate
Primary use casesText generation, Tool-use / function calling, Structured (JSON-style) output generation, Configurable reasoning tasks (high, low, no_think), Long-context processing up to 262,144 tokens, Text Generation

What to verify before adopting

    Evolution and major updates

    Hy3 Preview timeline

    A concise history of software releases and material product changes. Events appear only when a dated source supports what changed.

    1 dated update
    Latest first · exact dates

    Showing the newest updates and meaningful milestones. Open an entry for its summary and source.

    1. Release note

      Hugging Face repository published

      Open details

      The public model repository was created on Hugging Face. This repository date may differ from the model's original release date.

      View source [2]
    Citation ledger

    Recorded sources

    16 unique pages

    Facts, answers, structured details, milestones and primary resource links cite this shared ledger. Each external page appears once; release tags from the same GitHub project are grouped under one release history.

    1. 1huggingface.co/docs 12 facts · 3 answers · Documentation
    2. 2huggingface.co/tencent/Hy3-preview 9 facts · 1 milestone · Official site
    3. 3huggingface.co/inference/models 8 facts · 1 answer · Official site
    4. 4huggingface.co/models 4 facts · 1 answer · Official site
    5. 5huggingface.co/models 3 facts · 1 answer · Official site
    6. 6huggingface.co/blog Official site
    7. 7huggingface.co/docs/hub/eval-results Documentation
    8. 8huggingface.co/docs/inference-providers/index Documentation
    9. 9huggingface.co/docs/safetensors/index Documentation
    10. 10huggingface.co/models Official site
    11. 11huggingface.co/models Official site
    12. 12huggingface.co/models Official site
    13. 13huggingface.co/models Official site
    14. 14huggingface.co/pricing Pricing
    15. 15huggingface.co/privacy Security
    16. 16huggingface.co/support Official site
    Research status35 substantive facts · 16 source pages · quality score 88/100