Hy4 Flash
Overview
Hy4 Flash is Tencent's lightweight Hy4-series model for everyday productivity and high-volume workloads. It combines text and image understanding with a 1M-token context window, deep reasoning, and tool use.
Key Features
- Everyday Productivity: Designed for high-volume office tasks such as drafting, summarization, and information processing.
- Long-Context Analysis: Works with extensive documents, reference material, and conversation history within a 1M-token context window.
- Visual Understanding: Reads text in images, interprets charts, and analyzes screenshots alongside written instructions.
- Flexible Reasoning: Enables deep thinking by default and allows it to be disabled for simpler requests.
- Connected Workflows: Supports function calling, preserved thinking, and structured output for tasks that combine reasoning with external tools.
Best Use Cases
- Office Assistance: Draft correspondence, summarize reports, and organize information across routine business tasks.
- Document and Chart Analysis: Extract information from scanned pages, interpret charts, and compare images with written material.
- Coding Assistance: Explain code, investigate bugs, and support code generation or refactoring with relevant project context.
- Workflow Automation: Use connected tools to retrieve information, process results, and prepare structured responses across multiple steps.
Capabilities and Limitations
| Capability | Description |
|---|---|
| Input and Output | Text and image input; text output, including code and structured responses. |
| Reasoning | Deep thinking is enabled by default and can be disabled; supports preserved thinking across tool interactions. |
| Context Window | 1M tokens. |
| Maximum Input | 960,000 tokens. |
| Maximum Output | 64,000 tokens. |
| Tool Use | Function calling for external tools and multi-step workflows. |
| Structured Output | Supports structured responses for extraction and downstream processing. |
| Image Input | PNG, JPEG, WEBP, and static GIF; up to 10 MB per image and 600 images per request, including conversation history, subject to token limits. |
Known Limitations
- Images are internally resized to approximately 2048 x 2048 pixels, and inputs exceeding 5500 x 5500 pixels are rejected; crop relevant regions when small text or fine detail matters.
- Executing code, editing files, or operating applications requires an environment with the appropriate external tools and permissions.
- For best results, tool workflows using preserved thinking should retain the model's unmodified reasoning history between requests.
Standard Pricing
The token prices below are shown in USD per 1 million tokens; web search is billed in USD per use.
| Model | Input (USD / 1M Tokens) | Cache Write (USD / 1M Tokens) | Cache Read (USD / 1M Tokens) | Output (USD / 1M Tokens) | Web Search (USD / use) |
|---|---|---|---|---|---|
| Hy4 Flash | $0.07 | $0.07 | $0.003 | $0.281 | - |
Cache writes: B.AI bills cache writes at the input rate (1x).
Credits settlement: B.AI converts charges at 1 USD = 1,000,000 Credits and deducts Credits from the account balance.
Model prices in the documentation are based on the model providers' official list prices and serve as B.AI standard reference prices. B.AI may offer discounts or limited-time promotions on this basis; offer details are shown in Promotions and Pricing Updates and the website's discount groups. Applicable prices, offer terms, and final charges are subject to the platform display and actual billing records.