Compared With Hy4
Hy4 flash joined Tencent Hy on November 16, 2026. It is the lower-cost sibling of the Hy4 flagship. TokenHub lists both with text and image inputs, function calling, structured outputs and a 1M-token context window. The published limits are 960K input and 64K output tokens. TokenHub release log · Model limits
Start with Flash for recurring tasks with clear acceptance criteria, such as extracting fields from scanned forms, summarizing templated documents or classifying screenshots. Keep Hy4 as the baseline for long-running workflows with many tool calls. Before routing a workload to Flash, compare task completion, schema validity and retries on representative requests. Extra calls or tokens can change the cost per completed task.
Hy4 flash vs. Hy4 API Prices
International TokenHub list prices in USD per million tokens:
| Token type | Hy4 flash | Hy4 |
|---|---|---|
| Input | $0.070 | $0.834 |
| Output | $0.281 | $2.501 |
| Cached input | $0.003 | $0.042 |
The list rate is about 92% lower for input and 89% lower for output. In mainland China, the Flash list rates are CNY 0.50 for input, CNY 2.00 for output and CNY 0.02 for cached input per million tokens. TokenHub lists a 50% Flash inference discount through November 30, 2026 at 12:00 Beijing time. International rates · Mainland China rates
Access through TokenHub
Set the model ID to hy4-flash in Tencent Cloud TokenHub. Its international API docs list OpenAI Chat Completions, OpenAI Responses and Anthropic-compatible Messages. Use the endpoint and key for your TokenHub region. TokenHub API overview
Based on Tencent's public documentation. ToolWorthy did not run a hands-on comparison for this update.



