Tencent details Hy4 flash pricing, capabilities and availability
Tencent says Hy4 flash is its most cost-effective Hy model to date, with multimodal inputs and a context length of up to one million tokens. The date quoted by Tencent is November 16, 2026.

Key takeaways · 4
- 01
Hy4 flash supports multimodal inputs and a context length of up to one million tokens.
- 02
Its listed input and output rates are about 92% and 89% lower than Hy4’s, respectively.
- 03
Tencent recommends Hy4 flash for everyday tasks and full Hy4 for complex work such as long-horizon software engineering.
- 04
Tencent says the model is available through Tencent Cloud TokenHub and several other services; OpenRouter integration is planned.
A lower-priced option in Tencent’s Hy lineup
Tencent describes Hy4 flash as its most cost-effective Hy model to date.[1] Its listed API price is about $0.070 per million input tokens and $0.281 per million output tokens, compared with $0.834 and $2.501 for Hy4.[2] Tencent lists cached input at $0.003 per million tokens for Hy4 flash, versus $0.042 for Hy4.[2] The listed input and output rates are about 92% and 89% lower than Hy4’s, respectively.[2] Those differences make the published rates a practical starting point for teams comparing model costs, though actual spending depends on their usage.
Capacity and efficiency claims
Tencent says Hy4 flash accepts multimodal inputs and has a context length of up to one million tokens.[1] It has 224 billion total parameters, including 64 billion Engram parameters, and activates 12 billion parameters.[1] Tencent also says its KV cache is one twenty-fourth the size of Hy3’s and one-sixth of Hy4’s.[1] The company attributes decoding speeds above 150 tokens per second to megakernel optimization, four-bit quantization and DFly.[1] It says reinforcement learning reduced the model’s average end-to-end task output tokens by 50%.[1] These are Tencent’s stated specifications and optimization results.
Tencent’s evaluation and comparisons
Tencent says it evaluates deliverable quality, single-turn output length and task turnaround time rather than relying on public benchmark scores.[1] Its WorkIndex comprises eight internal evaluation sets covering work such as slide creation and editing, document editing, spreadsheet repair, audio and video processing, and visual understanding.[1] On that benchmark, Tencent reports Hy4 flash’s cost per task at one-thirteenth of Hy4 Preview’s, one-twentieth of Hy4’s and one-eighth of DeepSeek-V4.1-Flash-1017’s at Max-tier peak-hour pricing.[1] Tencent also says Hy4 flash slightly outperforms Hy4 Preview overall, although Preview has 770 billion total parameters and 49 billion active parameters.[1]
Availability and recommended use
Tencent says Hy4 flash is available through Tencent Cloud TokenHub, WorkBuddy, Yuanbao, WeCom, QQ, ima and Marvis, with OpenRouter integration planned.[1] The company offered free access on WorkBuddy through November 30 and a 50% discount on standard pricing through Tencent Cloud TokenHub during the promotion.[1] Tencent recommends the model for everyday work including deep research, document processing, PowerPoint generation, simple frontend displays and computer use.[1] It recommends full Hy4 for complex work, including long-horizon software engineering, saying that model is more capable.[1] Tencent says it plans to open-source Hy4 flash and update it frequently.[1]
For teams weighing AI model costs against task requirements, the listed price gap and Tencent’s everyday-versus-complex-work guidance offer concrete inputs for a pilot. The performance comparisons are company-reported and based on internal evaluations, so professionals should treat them as one source of evidence when selecting a model.
Why it matters
AI training for your team
kju.ai is built for teams: daily sessions, team analytics, and rollout support for your organization.
How this developed
19 November 2026
Tencent details Hy4 flash pricing, capabilities and availability