← All tools

Large Language Model

Hy4 Flash

Visit website ↗

About Hy4 Flash

Tencent’s Hunyuan team develops Hy4 Flash, an open-weight multimodal model intended to deliver capable results at comparatively low inference cost. It accepts text and images, generates text, and supports a context window of up to one million tokens; its mixture-of-experts architecture activates 12 billion parameters per token. The team recommends it for everyday productivity work such as deep search, document processing, slide creation, simple frontend demos, and computer use. Organizations can run it locally or deploy it with serving tools such as vLLM and SGLang, using an OpenAI-compatible API.

For business teams, Hy4 Flash can help summarize long reports, prepare presentation drafts, extract insights from visual materials, and support internal assistants or automated document workflows. Tencent reports performance above its Hy4 preview model on its WorkIndex evaluation at a substantially lower average task cost, though performance and costs will depend on deployment hardware and workload. Instruct, FP8, and base checkpoints are available, with fine-tuning support for teams adapting the model to specialized tasks. The model is released under the Tencent Hy License, so businesses should review its terms before commercial deployment.

What you can use it for

  • Summarize and process long business documents
  • Draft slide presentations and support slide editing
  • Search and analyze text and image inputs
  • Build internal productivity assistants through an OpenAI-compatible API
  • Create simple frontend prototypes
  • Fine-tune an open-weight model for specialized business workflows