Skip to content

Hy3 preview

Verified

Tencent's Hy3 preview manages 256k-token contexts for long-form text tasks.

TencentLanguage ModelsClosed
Model page Updated 2026-06-14

About Hy3 preview

Hy3 preview was developed as a proprietary LLM by Tencent. Its design centers on a 262144-token context length that supports lengthy input sequences. This architecture enables sustained coherence across very long documents without external retrieval mechanisms.

The model's strengths include reliable processing of extended text passages for analysis and synthesis. It suits professional workflows where maintaining context over hundreds of thousands of tokens is essential. Typical usage covers legal document review, technical report summarization, and multi-turn conversational agents with rich history.

Capabilities

Long-context reasoning
Document-level text analysis
Extended conversation handling
Text generation and summarization
Instruction following over large inputs

Best for

Long Document Analysis

Processes and reasons over full-length reports or books within its 262144-token context for deep insights.

Extended Conversation Management

Sustains coherent multi-turn dialogues across large histories while tracking prior details accurately.

Large Input Summarization

Generates summaries and follows complex instructions on extensive texts without losing key information.

Strengths & limitations

Strengths

  • +Very large 262k token context window
  • +Focused text-only processing
  • +Suitable for lengthy document tasks

Limitations

  • Text modality only with no vision support
  • Preview version may have variable stability

Where to access Hy3 preview

Frequently asked questions

Hy3 preview supports a context window of 262144 tokens.

Similar models

Other language models worth comparing.