Hy4 preview

tencent/hy4-preview-20260827

Tencent's Hy4 preview is a text-to-text AI model with a 1,048,576 token context length. It supports tool use and reasoning capabilities. The model is available as open weights, with a scheduled release date of August 28, 2026.

  • tool use
  • structured outputs
  • reasoning
  • open-weight

At a glance

providertencent
releasedAug 28, 2026
context1.0M
input / M$0.83

Specifications

What the API gives you

Context window
1.0M tokens
Max output
64K tokens
Input price
$0.83 / 1M tokens
Output price
$2.50 / 1M tokens
Cache read
$0.042 / 1M tokens
Modalities
text → text
Knowledge cutoff
—
Released
Aug 28, 2026

Measured evidence

Public benchmark record

LMArena Agent
2.52026-10-19
LMArena Agent Bash Recovery
6.52026-10-08
LMArena Agent Steerability
1.72026-10-19
LMArena Agent Task Outcome
5.72026-10-19
LMArena Agent Tool Hallucination
0.12026-10-19
LMArena WebDev
1631 Elo2026-10-19

LMArena data licensed CC-BY-4.0. Aider (Apache-2.0) and SWE-bench (MIT) are open. Design Arena uses blind human preference.

Open-weight model on Hugging Face.

Task evidence

How broad is the support?

Each status is exact to this model version and the selected task. Official release claims remain provisional.

  • Best LLM for Tool & Function CallingProvisional evidence

    38% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from Berkeley Function Calling or LMArena Search. 6 of 7 configured evaluation feeds currently have results.

  • Best LLM for Code CompletionProvisional evidence

    36% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from Aider Polyglot or LiveBench Coding. 3 of 3 configured evaluation feeds currently have results.

  • Best LLM for Hermes AgentProvisional evidence

    29% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from Berkeley Function Calling or SWE-rebench. 4 of 4 configured evaluation feeds currently have results.

  • Best LLM for OpenHandsProvisional evidence

    29% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from Berkeley Function Calling or SWE-rebench. 4 of 4 configured evaluation feeds currently have results.

  • Best LLM for GooseProvisional evidence

    29% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from Berkeley Function Calling or SWE-rebench. 4 of 4 configured evaluation feeds currently have results.

  • Best LLM for OpenClawProvisional evidence

    29% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from Berkeley Function Calling or SWE-rebench. 4 of 4 configured evaluation feeds currently have results.

Show all 14 tasks
  • Best LLM for Frontend & UIProvisional evidence

    27% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from Design Arena Coding or Design Arena UI. 6 of 6 configured evaluation feeds currently have results.

  • Best LLM for AI AgentsProvisional evidence

    25% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from Berkeley Function Calling or LiveBench Agentic Coding. 9 of 9 configured evaluation feeds currently have results.

  • Best LLM for Autonomous AgentsProvisional evidence

    22% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from LiveBench Agentic Coding or OSWorld Verified. 6 of 6 configured evaluation feeds currently have results.

  • Best Local LLM for CodingProvisional evidence

    21% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from Aider Polyglot or LiveBench Coding. 4 of 4 configured evaluation feeds currently have results.

  • Best LLM for CodingProvisional evidence

    20% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from Aider Polyglot or Design Arena Coding. 7 of 7 configured evaluation feeds currently have results.

  • Best LLM for Agentic CodingProvisional evidence

    19% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from Design Arena Full-stack or Design Arena Web-apps. 7 of 7 configured evaluation feeds currently have results.

  • Best LLM for Debugging & Code ReviewProvisional evidence

    17% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from Aider Polyglot or LiveBench Coding. 5 of 5 configured evaluation feeds currently have results.

  • Best LLM for Computer & Browser UseProvisional evidence

    17% of intended weighted task evidence across 1 independent source; broader independent confirmation is still missing from Berkeley Function Calling or OSWorld Verified. 3 of 3 configured evaluation feeds currently have results.

Recommendation coverage

Where this model appears

Not currently included in a published recommendation. The task evidence panel above says exactly what is confirmed and what is still missing.

Catalog data via OpenRouter and models.dev. Last checked within the past 6 hours.