Hy4

Tencent's Hy4 LLM family: 770B-parameter (49B active) MoE with 1M context; full release adds visual understanding, with Hy4 Flash for cost.

Tencent · Frontier/foundation models · model

Published in Feed

This entry is marked for review. Check the original sources before relying on its details.

Overview

Hy4 is Tencent's large language model family. An open Hy4 preview (770B total, 49B active parameters, 1M-token context) appeared on Hugging Face on August 28, 2026. At the Tencent Global Digital Ecosystem Summit on October 29, 2026, Tencent said the full Hy4 and a cheaper Hy4 Flash would follow soon with first-time visual understanding and better long-horizon agentic performance; a tencent/hy4 listing on OpenRouter (Tencent Cloud pricing $0.417 per million input and $1.251 per million output tokens) and a release tracker dated November 4 indicate the full model has shipped. Open weights for the full release are unconfirmed.

This is a sourced directory entry. Features and capabilities are reported in the linked sources; this entry does not document independent Aymorphic testing.

Key features

  • Preview: 770B total / 49B active parameters, 1M-token context
  • Full release adds visual understanding, per Tencent
  • Hy4 Flash targets cost efficiency while keeping vision and agent skills
  • OpenRouter listing supports streaming, tools, structured output and reasoning parameters
  • Tencent Cloud API pricing $0.417 / $1.251 per million input / output tokens (OpenRouter)

Use cases

  • Coding agents and tool-use workflows
  • Long-horizon productivity tasks over long documents
  • Image-grounded reasoning in agents

Pricing

Via OpenRouter / Tencent Cloud: $0.417 per million input tokens and $1.251 per million output tokens; Hy4 Flash pricing unknown.

Pricing reflects the source record and may change. Confirm current terms with the provider.

Dates and sources

Released
4 Nov 2026
Record updated
6 Nov 2026
Record last checked
6 Nov 2026

Explore the Feed

Other entries filed under frontier/foundation models; inclusion does not imply equivalent features.

  • Muse Wonder

    Meta's most capable Muse models, with active memory and long-goal focus for agents, 1M context and $4/$20 per million tokens.

  • Rho-1

    Reka's 19B omni-reasoning model that handles text, images, video and robot actions in one network, in research preview.

  • Amazon Nova 3

    Amazon's third-generation Nova frontier model family, unveiled at an AWS keynote with a 1M-token context window and a dedicated agentic tool-use variant.

Browse the Feed