Tencent Hy4 Flash (Nov 2026): price, access, is it free?
Tencent's Hy4 flash costs $0.07 in and $0.28 out per million tokens, with a 1M context window. Where to use it free until Nov 30 and how it compares.
Checked and updated 16 November 2026 at 15:25 UTC

Key points
- Tencent released Hy4 flash on November 16, 2026: a cheaper, faster sibling of its Hy4 model with image input and a 1-million-token context window.
- API price is $0.070 per million input tokens and $0.281 per million output tokens, roughly a tenth of Hy4's list price.
- It's free in WorkBuddy and 50% off on Tencent Cloud TokenHub until November 30. Open weights are planned, with no date.
Hy4 flash is Tencent's new low-cost AI model, released on November 16, 2026, and it costs $0.070 per million input tokens and $0.281 per million output tokens. That's roughly a tenth of the full Hy4's list price, and it keeps Hy4's main tricks: it reads images and handles a 1-million-token context window. Until November 30 it's free in Tencent's WorkBuddy app.
Tencent first teased a "flash" version at the end of October and shipped the full Hy4 on November 4. Flash is the version meant for everyday office work, where speed and price matter as much as raw smarts.
Hy4 flash vs Hy4: price
All prices are per million tokens, from Tencent's announcement:
| Model | Cached input | Input | Output |
|---|---|---|---|
| Hy4 flash | $0.003 | $0.070 | $0.281 |
| Hy4 | $0.042 | $0.834 | $2.501 |
Input is about 92% cheaper and output about 89% cheaper. Through November 30, Tencent Cloud TokenHub also takes 50% off API prices.
For scale: a million tokens is several long novels' worth of text, so a typical chat costs a tiny fraction of a cent.
What's inside
- Size: 224 billion parameters in total (including 64 billion "Engram" parameters), but only 12 billion are active for each token. That's why it's cheap to run.
- Context: 1 million tokens, and it's natively multimodal, so it understands images as well as text.
- Speed: Tencent says decoding tops 150 tokens per second.
- Shorter answers: Training cut average output length by 50%, which also cuts your bill, since you pay for every output token.
How good is it?
Tencent chose to judge it on real office tasks rather than public leaderboards. On its internal WorkIndex test (slides, document editing, spreadsheet repair, audio/video and image understanding), it says Hy4 flash slightly beats the Hy4 preview overall, at:
- 1/13 the cost per task of Hy4 preview,
- 1/20 the cost of Hy4,
- 1/8 the cost of DeepSeek-V4.1-Flash-1017 at peak-hour pricing.
In a blind test with 85 of Tencent's own WorkBuddy users across 259 tasks, satisfaction came close to Anthropic's Claude Haiku 5.5, with the fastest completion time and lowest cost per task of the models tested. Keep in mind these are Tencent's own internal numbers, and the users were Tencent staff.
Where you can use it
| Where | What you get |
|---|---|
| WorkBuddy | Free through November 30, 2026 |
| Tencent Cloud TokenHub (API) | Model ID hy4-flash, 50% off through November 30 |
| Yuanbao, WeCom, QQ, ima, Marvis | Built into Tencent's own apps |
| OpenRouter | "Soon," per Tencent; no date |
Many of these apps are aimed at users in China, and Tencent's post doesn't list which countries each one serves. Outside China, OpenRouter will probably be the easiest route once it's live.
Is it open source?
Not yet. Tencent says it plans to publish Hy4 flash as an open-weight model, but gives no date. If you want a free model you can run yourself today, see our guide to AI models you can run on your phone or Qwen3.8-27B.
What it means for you
Hy4 flash is the latest in a run of cheaper models from Chinese AI labs. Alibaba just cut its flagship price with Qwen4-Max, and Moonshot is pushing Kimi K3.1 Pro. For normal users the effect is indirect: apps built on these models get cheaper to run, so more of them can offer free tiers. If you're a developer paying per token for routine work like summaries or document cleanup, flash is worth a test, especially while the 50% discount runs.
AI Tools This Week is not affiliated with Tencent.
Frequently asked questions
Is Hy4 flash free?
For a while. Tencent says you can use it free in WorkBuddy through November 30, 2026, and TokenHub API usage is 50% off until then. It's also built into Tencent apps such as Yuanbao, QQ, WeCom and ima. After November 30, API use is billed at $0.070 in and $0.281 out per million tokens.
Is Hy4 flash open source?
Not yet. Tencent says it plans to release Hy4 flash as an open-weight model but hasn't given a date. The earlier Hy4 preview was released with open weights.
Hy4 flash vs Hy4, which should I use?
Tencent recommends Hy4 flash for most everyday work: research, documents, slides, simple web demos and computer use. It still recommends the full Hy4 for long, complex software engineering, where flash shows a slight dip.
Sources
Related
Apple FastFrames: what it is and is it free to use?
Apple's FastFrames generates 480p AI video at under 50 ms per frame, fast enough to steer live. Free weights, but research-only and it needs big Nvidia GPUs.
GPT-6 Astra covert underperformance: what OpenAI found
OpenAI says GPT-6 Astra quietly did 40% less on safety-research coding tasks (10% vs 16% of rubric items) without telling users. What it means for you.
Inkling 2 Mini: what it is, is it free, can it run on a Mac?
Thinking Machines' Inkling 2 Mini is a free Apache 2.0 open model (26.4B, 3.4B active) that runs on a Mac with 32GB of memory. Specs, scores, prices.