Tencent Launches Hy4 Flash: A Cheaper AI Model for Everyday Work at $0.07 Per Million Tokens
Key takeaway: Tencent launched Hy4 Flash on November 16, 2026, a smaller, cheaper sibling of its Hy4 flagship AI model. It has 224 billion parameters, a one-million-token context window, and API prices starting at $0.07 per million input tokens — roughly a twelfth of the full Hy4’s list price. It is aimed at everyday office work, not frontier research.
What Tencent announced
Tencent launched Hy4 Flash on Monday, November 16, 2026, less than two weeks after introducing the full Hy4 flagship on November 4. Hy4 Flash is the cost-efficient tier Tencent promised at that launch. TechNode reported the launch the same day, and Tencent Cloud’s updated documentation lists the new model, its API prices, and its specifications.
Hy4 Flash is a mixture-of-experts model with 224 billion total parameters, of which 12 billion activate for each token it processes. It accepts multimodal input — text and images — and has a one-million-token context window, with a maximum input of 960,000 tokens and maximum output of 64,000 tokens, according to Tencent Cloud’s model list. It supports image understanding, structured output, function calling, caching, and deep reasoning.
How much Hy4 Flash costs
The price is the headline. Tencent Cloud lists Hy4 Flash at $0.07 per million input tokens, $0.281 per million output tokens, and $0.003 per million cached-input tokens. The full Hy4 lists at $0.834, $2.501, and $0.042 — roughly twelve times more on input and nine times more on output.
| API price (per 1M tokens) | Hy4 Flash | Hy4 (full) |
|---|---|---|
| Input | $0.07 | $0.834 |
| Output | $0.281 | $2.501 |
| Cached input | $0.003 | $0.042 |
Tencent says that on its internal WorkIndex benchmark, Hy4 Flash completed everyday tasks — research, slide creation, and spreadsheet editing — at one-twentieth the per-task cost of the full Hy4, or about 95% less. That figure is Tencent’s own claim from its own internal benchmark, not an independent evaluation. WorkIndex is built from typical WorkBuddy office tasks and is not public.
Two caveats narrow the gap in practice. The full Hy4 is itself on a 50% discount, and Hy4 Flash is free in WorkBuddy through November 30, with TokenHub at half price until then — launch promotions, not permanent prices.
Hy4 Flash vs Hy4: which one fits what
The two models share the same one-million-token context window and the same listed input and output limits, so context capacity alone is a poor reason to pick one over the other. The difference is scale and cost: Hy4 has 770 billion total parameters with 49 billion active per token, plus native vision and computer-use capabilities. Hy4 Flash has 224 billion total with 12 billion active, and is positioned for routine, high-volume work — documents, slides, spreadsheets, and deep search.
The pattern is familiar across the industry. Meta’s Muse Ember, launched earlier in November, and Google’s Gemini 4 Argon, launched in October, both target the same idea: a cheaper tier for work that does not need a flagship model.
Where you can use it
Hy4 Flash is live now through Tencent Cloud’s TokenHub and inside Tencent’s own products: WorkBuddy, Yuanbao, WeCom, QQ, ima, and Marvis. Tencent says OpenRouter support is next. The API guide lists support for OpenAI Chat Completions, OpenAI Responses, and Anthropic Messages protocols. Tencent also says it plans to release Hy4 Flash’s model weights, but has given no date. Until the weights arrive, the model is available only as a hosted service.
Knowns, unknowns, and what to watch
Known: the launch date, the parameter counts, the one-million-token context, and the published API prices. Unknown: how Hy4 Flash performs on independent benchmarks, and when — or whether — the open weights arrive. What to watch: OpenRouter support, independent leaderboard results, and whether the November 30 promotions get extended.
Analysis: the cheap tier is the real product
Analysis — this section is our assessment, not a verified fact. Flagship model launches get the headlines, but the cheaper tier is where most real usage lands. A model that costs a twelfth of the flagship per token changes which tasks are worth automating at all: document-heavy workflows, long-context summarisation, and agent loops that make dozens of calls per job only make economic sense at this price point. The open question is quality: a 95% cost cut means little if the cheaper model needs more retries or more human review. Until independent benchmarks land, the honest move is a small, bounded trial on your own documents — measuring completed work and review time rather than trusting the price list or the vendor benchmark. WorkBuddy users can try it free until November 30.
Frequently asked questions
What is Tencent Hy4 Flash?
Hy4 Flash is a lower-cost AI model Tencent launched on November 16, 2026. It is a 224-billion-parameter mixture-of-experts model with 12 billion parameters active per token, multimodal input, and a one-million-token context window. It is the cheaper sibling of the Hy4 flagship, aimed at everyday office work.
How much does Hy4 Flash cost?
Tencent Cloud lists Hy4 Flash at $0.07 per million input tokens, $0.281 per million output tokens, and $0.003 per million cached-input tokens — roughly a twelfth of the full Hy4 model’s list price. Through November 30, 2026, it is free in WorkBuddy and half price on TokenHub.
How does Hy4 Flash compare to the full Hy4 model?
Both models have a one-million-token context window and the same listed input and output limits. The full Hy4 is much larger — 770 billion total parameters with 49 billion active — and adds native vision and computer-use capabilities. Hy4 Flash is smaller and far cheaper, built for routine, high-volume tasks.
Where can I use Hy4 Flash?
Hy4 Flash is available through Tencent Cloud’s TokenHub and inside Tencent products including WorkBuddy, Yuanbao, WeCom, QQ, ima, and Marvis. Tencent says OpenRouter support is coming next, and it plans to release the model’s weights, though no date has been given.
Sources: TechNode; CASRAI, citing Tencent Cloud’s API guide, model list, and pricing table; AI Frontier Post.
One Comment
Comments are closed.