Tencent Hy4 flash: price, specs and how to try it.
Tencent’s Hy4 flash is a 224B MoE model with 12B active parameters and a 1M context at about $0.07 per million input tokens. Specs, access and use cases.
By The Model Atlas · 4 sources
In this post7 sections
Tencent Hy4 flash is the newest and, Tencent says, most cost-effective model in its Hy family. Announced on 16 November 2026, it pairs 224 billion total parameters with only 12 billion active per token, takes multimodal input, handles up to 1M tokens of context, and Tencent prices it at about $0.07 per million input tokens. Here is what Hy4 flash is, what Tencent says it does well, what it costs and how to try it today.
What is Tencent Hy4 flash?
Hy4 flash is a smaller, faster sibling of Hunyuan Hy4, the flagship Tencent released on 4 November. Tencent says Hy4 flash has 224 billion total parameters, including 64 billion Engram parameters, with 12 billion active, supports multimodal inputs and a context length of up to 1M tokens, and is “the most cost-effective Hy model to date”.
Tencent pitches it for everyday work. The team recommends Hy4 flash for most daily tasks, including deep research, document processing, PowerPoint deck generation, simple front-end displays and computer use, and points users to the larger Hy4 for complex jobs such as long-horizon software engineering.
What’s new in Hy4 flash
According to Tencent’s announcement, the gains come from choices across the whole pipeline:
- Architecture. Hy4 flash uses hybrid attention, LatentMoE and Engram features. Tencent says its KV cache is 1/24 the size of Hy3’s and 1/6 the size of Hy4’s.
- Shorter answers. Tencent says reinforcement learning cuts average end-to-end task output tokens by 50 percent.
- Speed. With megakernel optimization, 4-bit quantization and DFly techniques, Tencent says decoding exceeds 150 tokens per second.
- Work-focused testing. The team built WorkIndex, an internal benchmark based on typical WorkBuddy tasks, scored on deliverable quality, output length and turnaround time rather than public benchmark scores alone.
Hy4 flash price
Tencent lists approximately these API prices per million tokens:
- Input: $0.070
- Output: $0.281
- Cached input: $0.003
Tencent calls this a substantial reduction from Hy4. It also says that until 30 November users can use Hy4 flash for free in WorkBuddy and get 50 percent off standard pricing on Tencent Cloud TokenHub.
Hy4 flash benchmarks and performance
Tencent’s headline claims are about cost per task rather than leaderboard scores:
- Versus Hy4 Preview. Tencent says Hy4 flash slightly outperforms the larger Hy4 Preview (770 billion total, 49 billion active parameters) in overall effectiveness.
- Cost per task. On WorkIndex, which covers slide creation and editing, document editing, spreadsheet repair, audio and video processing and visual understanding, Tencent says Hy4 flash costs 1/13 as much per task as Hy4 Preview, 1/20 as much as Hy4 and 1/8 as much as DeepSeek-V4.1-Flash-1017 (Max tier, peak-hour pricing).
- Blind test. In a blind evaluation with 85 internal WorkBuddy users across 259 tasks in 27 categories, Tencent says Hy4 flash reached user satisfaction close to Claude Haiku 5.5, with the lowest per-task cost and the shortest completion time of the models tested.
These are Tencent’s own internal results. Tencent says its comparison charts also include seven public agent benchmarks, but independent scores were not part of the announcement.
How to try Hy4 flash
Tencent says Hy4 flash is available now in Tencent Cloud TokenHub, WorkBuddy, Yuanbao, WeCom, QQ, ima and Marvis, with OpenRouter integration coming soon. To call it through the API:
- Create a Tencent Cloud TokenHub API key and activate the Hy4 flash model service in the console. Tencent’s docs say calls to a model that is not activated return a 402 error.
- Point an OpenAI- or Anthropic-compatible client at TokenHub. For the global region, Tencent’s docs list the Singapore base URL `https://tokenhub-intl.tencentcloudmaas.com/v1`.
- Set the model to `hy4-flash`. Tencent’s docs list support for OpenAI Chat Completions, OpenAI Responses and Anthropic Messages.
If you just want to try it without code, WorkBuddy is the simplest route, and it is free there until 30 November according to Tencent.
What people are building with Hy4
Hy4 flash only launched on 16 November, so there are no Hy4 flash builds on The Model Atlas yet. Builders have been busy with its big sibling, Hy4, though, and those projects show the kind of work Tencent is now pitching Flash for:
- A playable 3D browser game from one prompt that you can walk, jump and interact with.
- Neon Drift, a browser racing game with drifting, nitro and AI rivals, made in WorkBuddy (the post is marked as an ad).
- An animated story site for an original game, also built inside WorkBuddy.
- A small observatory game built in about an hour.
New Hy4 flash builds will appear on the Hy4 flash model page as creators share them.
Who is Hy4 flash for?
Going by Tencent’s description, Hy4 flash is for teams and developers running high-volume everyday work, such as research, documents, slide decks, spreadsheets, simple front ends and computer-use agents, who want a long context and multimodal input at a very low price per task. For long, complex coding projects, Tencent itself recommends the larger Hy4.
When was Hy4 flash released?
Tencent announced Hy4 flash on 16 November 2026, twelve days after the flagship Hy4 launched on 4 November.
How much does Hy4 flash cost?
Tencent lists about $0.070 per million input tokens, $0.281 per million output tokens and $0.003 per million cached input tokens. Until 30 November it is free in WorkBuddy and 50 percent off on Tencent Cloud TokenHub.
How many parameters does Hy4 flash have?
Tencent says 224 billion in total, including 64 billion Engram parameters, with 12 billion active per token.
What is the Hy4 flash context window?
Up to 1M tokens, according to Tencent.
What is the Hy4 flash model ID?
`hy4-flash` on Tencent Cloud TokenHub.
Is Hy4 flash open source?
Not yet. Tencent says it plans to open-source Hy4 flash, without giving a date.
Is Hy4 flash on OpenRouter?
Tencent says OpenRouter integration is coming soon. For now, use Tencent Cloud TokenHub or Tencent’s apps such as WorkBuddy and Yuanbao.
Hy4 flash vs Hy4: which should I use?
Tencent recommends Hy4 flash for most everyday tasks and the larger Hy4 for complex work such as long-horizon software engineering.
Sources · 4
- Tencent Launches Hy4 flash to Power High-Speed, Cost-Efficient AI Deployment for Everyday Workflows (Tencent, 16 November 2026) (opens in a new tab)
- Language Model API Overview (Tencent Cloud TokenHub docs) (opens in a new tab)
- API Usage Instructions (Tencent Cloud TokenHub docs) (opens in a new tab)
- TokenHub (Tencent Cloud) (opens in a new tab)