Back home

Hy4 flash

x+web 0.95
Visit source

Intro

Hy4 flash is the lower-cost Tencent Hunyuan model formally launched on 16 November 2026, positioned beside the existing flagship Hy4 for everyday productivity. @TencentHunyuan posted the introduction at 2026-11-16T06:38:39Z (14:38:39 Beijing time), beginning “Meet Hy4 flash, the cheapest model for everyday productivity.” and pointing to the official page at hy.tencent.ai/research/hy4-flash. The post says it matches Claude Haiku 5.5 (high) on daily work in human blind tests, with about an 18× reduction in task cost, and lists API prices of $0.003 cached, $0.070 input, and $0.281 output per million tokens.

The official page is dated November 16, 2026 and titled “Introducing Hy4 flash: Ultimate cost-efficiency for everyday productivity.” It describes Hy4 flash as the most cost-effective Hy model to date: 224B total parameters (including 64B Engram parameters) with 12B active per token, natively multimodal like Hy4, and a 1M-token context window, at substantially lower API list prices. The same pricing table lists flagship Hy4 at $0.042 cached, $0.834 input, and $2.501 output per million tokens. Tencent says Hy4 flash is now available on Tencent Cloud TokenHub, WorkBuddy, Yuanbao, WeCom, QQ, ima, Marvis, and more, and will launch on OpenRouter soon; through 30 November it is free in WorkBuddy, with 50% off TokenHub API pricing. The page recommends Hy4 flash for most everyday productivity work (deep search, document processing, slide creation, simple frontend demos, computer use), and still recommends Hy4 for complex workloads such as long-horizon software engineering. Tencent also plans to release Hy4 flash as an open-weight model and to ship frequent updates.

A same-day TechNode report restates the official figures: on Tencent’s internal WorkIndex benchmark, Hy4 flash cost about one-twentieth as much per task as full Hy4; in a separate blind test, 85 WorkBuddy users rated 259 tasks, and Tencent said satisfaction was close to Claude Haiku 5.5, with the lowest cost and shortest completion time among the models tested. Report: https://technode.com/2026/11/16/tencent-launches-hy4-flash-cutting-ai-task-costs-by-95-in-internal-tests/

Key points:

Advertisement

Explain

Why it matters

This is a new model name Tencent Hunyuan confirmed on 16 November 2026 with an official account post and a product page, not a routine price cut or feature update to the existing flagship Hy4. When Hy4 itself launched on 4 November 2026, official materials already said “Hy4 Flash is scheduled for release soon.” Inside this 16:32 scan window, the official introduction post (06:38:39Z / 14:38:39 Beijing time) and product page present Hy4 flash as a distinct lower-cost tier, with its own parameter counts, list prices, and availability, so it is filed separately.

This page only restates the official introduction post, the official product page, and a same-day TechNode report of the official figures. Claims that it “matches Claude Haiku 5.5 (high)” or cuts task cost by about 18× / to about one-twentieth of Hy4 are company statements; this page does not independently reproduce them.

Limits

Open weights and OpenRouter availability are still described as planned — do not write them as already shipped. The WorkBuddy free access and TokenHub 50% discount run through 30 November; recheck the official page after that date. Do not treat this page as a guarantee of quality, speed, or price, and do not describe Hy4 flash as a replacement for flagship Hy4.

Usage

  1. Read the official introduction at https://x.com/TencentHunyuan/status/2122217770314301469 and cross-check the official page at https://hy.tencent.ai/research/hy4-flash ; a same-day report is at https://technode.com/2026/11/16/tencent-launches-hy4-flash-cutting-ai-task-costs-by-95-in-internal-tests/ ;
  2. Follow the official instructions to try it on Tencent Cloud TokenHub, WorkBuddy, Yuanbao, WeCom, QQ, ima, or Marvis; the WorkBuddy free access and TokenHub 50% discount through 30 November are as stated on the official page; wait for a further official notice before assuming OpenRouter or open weights are live;
  3. When citing 224B / 12B, the 1M context window, per-million-token list prices, an ~18× task-cost reduction, or the Claude Haiku 5.5 comparison, attribute them to the 16 November 2026 official post, product page, or reporting of those official figures, and recheck the upstream source;
  4. Do not describe Hy4 flash as the 4 November 2026 flagship Hy4, or as the existing term Hy Realtime; this page only restates official materials and reporting — recheck the upstream sources before citing them externally.