█

AI/TLDR

Buy me a coffee
Buy me a coffee

Hunyuan Hy4

The full release of Tencent's Hy4, adding image input and computer use on a 1M-token context, released November 4, 2026.

IN SHORT

What is Hunyuan Hy4?

Hunyuan Hy4 is the full release of the Hy4 generation from Tencent's Hy Team, announced on 4 November 2026 after the open-weight Hy4 preview of August 2026. Tencent substantially scaled both pre-training and post-training over the preview and reports gains in instruction following, long-context understanding, long-horizon tasks and deep research. It has a 1M-token context window and costs $0.834 per million input tokens and $2.501 per million output tokens.

Older version
Hunyuan Hy4 preview · 28 August 2026
Where to use it
OpenRouter
Updated Checked against 2 sourcesMarkdown version ↗
Cite this pageAI/TLDR, "Hunyuan Hy4: specs, pricing and benchmarks", https://ai-tldr.dev/models/hunyuan-hy4/ (updated 4 November 2026)
// FACTSas of 4 November 2026
Context window1Mmax output 65,536 tokens[1]
Input price$0.834 / 1M tokenscached $0.042[2]
Output price$2.501 / 1M tokens[2]
LicenceProprietary (hosted API) · API only[1]
Released4 November 2026[1]
ParametersUndisclosed[1]
  1. [1] openrouter.ai OpenRouter — Tencent: Hy4
  2. [2] unite.ai Unite.AI — Tencent Releases Full Hy4 Model With Vision and Computer Use
HunyuanAPI onlyGenerally available — hosted through Tencent Cloud and OpenRouter and inside Tencent products. Tencent plans to publish the weights later with a comprehensive system card.
Released
4 Nov 2026
Context
1M
Parameters
Undisclosed
Input
$0.834 / 1M tokens
License
Proprietary (hosted API)
Coverage
1 story

Overview

Hunyuan Hy4 is the full release of the Hy4 generation from Tencent's Hy Team, announced on 4 November 2026 after the open-weight Hy4 preview of August 2026. Tencent substantially scaled both pre-training and post-training over the preview and reports gains in instruction following, long-context understanding, long-horizon tasks and deep research. The biggest functional change is multimodality: Hy4 reads fine image detail and complex charts, and it can operate computers across command-line and graphical interfaces.

The hosted model is listed on OpenRouter as tencent/hy4, served by Tencent Cloud, with a 1,048,576-token context window, a 65,536-token completion limit, text and image input and text output, and a reasoning mode with high, low and none effort levels. It is also built into Tencent's CodeBuddy, WorkBuddy, Yuanbao and ima products. Unlike the preview, the final weights are not out yet: Tencent says they will follow with a comprehensive system card and ecosystem support developed with the vLLM and SGLang communities, and that a faster, more cost-efficient Hy4 Flash is planned.

Tencent aimed the training at professional work. It covered Blender, CAD and long workflows that span several applications, front-end development, and game development in Unreal Engine, Unity and Godot, where Tencent reports gains of 19.1% on GameDev and 9.8% on GameLogic over the preview. In a company-run finance evaluation with 213 professionals across 579 real tasks, Tencent says Hy4 scored 3.04, ahead of DeepSeek V4.1 Flash 1017 at 2.97 and Kimi K3.1 Pro and GLM-5.3 at 2.68 each. Tencent also states that a gap remains to the top proprietary models on research-level mathematics, long-horizon coding, model self-improvement and advanced visual question answering.

Released2026-11-04
LicenseProprietary (hosted API)
WeightsAPI only
ParametersUndisclosed
Context1M
Max output65,536 tokens
ModalitiesText, Vision
StatusGenerally available — hosted through Tencent Cloud and OpenRouter and inside Tencent products. Tencent plans to publish the weights later with a comprehensive system card.

Pricing

Input$0.834 / 1M tokens
Cached input$0.042 / 1M tokens
Output$2.501 / 1M tokens

Tencent's listed API price, unchanged from Hy4 preview (as reported by Unite.AI). OpenRouter's live listing for tencent/hy4 currently shows $0.417 input and $1.251 output per 1M tokens.

Pricing source ↗

Strengths

  • Image input and computer use across command-line and graphical interfaces, new over the text-only Hy4 preview
  • 1,048,576-token context window with a 65,536-token completion limit on OpenRouter
  • Low hosted price: $0.834 input / $2.501 output per 1M tokens at Tencent's list price
  • Targeted training for Blender, CAD, front-end and game-engine work (Unreal Engine, Unity, Godot)
  • Tencent-reported drop in internal search hallucinations from 15.76% (preview) to 5.45%
  • Cybersecurity training data limited to defensive work, with red-team audits by Tencent Xuanwu Lab and Keen Security Lab

Best for

  • Computer-use agents that operate desktop and terminal applications
  • Chart, document and screenshot understanding inside long-context workflows
  • Game-development and 3D/CAD assistance
  • Finance analysis tasks over many files and workspaces
  • Low-cost hosted reasoning and coding through OpenRouter or Tencent Cloud

How to access

ProviderModel ID
OpenRouter ↗tencent/hy4

Hunyuan — every version

The full lineage of the Hunyuan line, newest first. Every version has its own page — click any to compare specs, benchmarks and pricing.

VersionReleasedContextLicense
Hunyuan Hy4current2026-11-041MProprietary (hosted API)
Hunyuan Hy4 preview2026-08-281MApache-2.0
Hunyuan Hy32026-07-06256KApache-2.0

FAQ

Is Hunyuan Hy4 open source?

Not yet. The full Hunyuan Hy4 is hosted only, through Tencent Cloud, OpenRouter and Tencent apps. Tencent says the weights will follow with a comprehensive system card and vLLM and SGLang support. The earlier Hy4 preview from August 2026 is open-weight on Hugging Face under Apache 2.0.

What is new in Hy4 compared with Hy4 preview?

Hunyuan Hy4 accepts images as well as text, where the preview was text-only, and adds computer use across command-line and graphical apps. Tencent scaled pre-training and post-training and reports gains in instruction following, long context and long-horizon tasks, plus 19.1% on GameDev and 9.8% on GameLogic over the preview.

How much does the Hy4 API cost?

Tencent's listed price for Hunyuan Hy4 is $0.834 per million input tokens, $2.501 per million output tokens and $0.042 per million cached input tokens, the same as the preview. OpenRouter's listing for tencent/hy4, served by Tencent Cloud, currently shows $0.417 per million input tokens and $1.251 per million output tokens.

What is Hy4 Flash?

Hy4 Flash is a separate model Tencent plans to release in the near future, described as a variant of Hy4 with much better cost efficiency and faster inference. Tencent has not given a release date, size or price for Hy4 Flash.