On this page
Following the November 4 release of Hy4, Tencent introduced Hy4 flash on November 16, 2026, establishing a two-tier lineup for everyday productivity versus deep reasoning. Developers selecting between the two should separate official workload routing, TokenHub's RMB catalog pricing and distinct 50% discount windows, and the USD reference pricing published on the research page.
Official Workload Routing & Technical Specifications
Hy4 flash for Everyday Productivity vs. Hy4 for Complex Engineering: On the Hunyuan research page, Tencent recommends Hy4 flash for most everyday productivity tasks—such as deep search, document processing, slide creation, simple frontend demos, and computer use (phrased in the press release as deep research, document processing, PowerPoint deck generation, simple front-end displays, and computer use)—while continuing to recommend Hy4 for complex workloads like long-horizon software engineering.
Architecture & Official Benchmark Disclosures: Hy4 flash features 224B total parameters (including 64B Engram) with 12B activated per token, a 1M context window, and decoding speeds exceeding 150 tokens/s. In Tencent's internal WorkIndex evaluations, average cost per task is roughly 1/13 that of Hy4 preview and 1/20 that of Hy4. Tencent also states it plans to release Hy4 flash as an open-weight model later (open-source weights are not live yet).
TokenHub RMB Catalog Pricing, Discount Windows & USD Reference Rates
On TokenHub's model list (across both Guangzhou and Singapore tabs), both hy4-flash and hy4 offer a 1M context window, 960k max input, 64k max output, preserved thinking, structured output, function calling, caching, and image understanding:
TokenHub RMB Catalog Rates (per 1M Tokens, Cache Writes Unlisted): Hy4 flash (hy4-flash) is listed at ¥0.50 input, ¥2.00 output, and ¥0.02 cache hit; both Hy4 (hy4) and Hy4 preview (hy4-preview) are listed at ¥6.00 input, ¥18.00 output, and ¥0.30 cache hit. Batch inference pricing lists Hy4 and Hy4 preview at ¥3.00 input, ¥9.00 output, and ¥0.15 cache hit (Hy4 flash does not appear in the batch table).
Two Separate 50% Discount Windows (Beijing Time): Hy4 flash calls are billed at 50% of catalog pricing through November 30, 2026 at 12:00:00 (Beijing Time), whereas Hy4 calls are billed at 50% of catalog pricing through December 31, 2026 at 23:59:59 (Beijing Time; the Hy4 discount notice specifies the Hy4 model and does not explicitly cover hy4-preview). Users can also access Hy4 flash for free in WorkBuddy through November 30.
Research Page USD Reference Table (per 1M Tokens): Separately from TokenHub RMB rates, the research page lists Hy4 flash at $0.070 input, $0.281 output, and $0.003 cached input, compared with $0.834 input, $2.501 output, and $0.042 cached input for Hy4.

Image source: compiled by DigIntoAI from Tencent Cloud TokenHub pricing docs, model list, and Announcement 2559.
Hy4 preview (hy4-preview) Retirement on December 7, 2026
Migration Rules for hy4-preview: Per Tencent Cloud Announcement 2559, only the preview model Hy4 preview (model ID hy4-preview) retires on December 7, 2026 at 00:00:00 Beijing Time across TokenHub, ADP, and CloudBase. Eligible unmigrated workloads will be automatically switched to Hy4 (hy4), while ineligible requests after retirement will return errors.
