Skip to content
All systems operational0 AI providers monitored, polled every 2 minutes
Live status

Hy4 flash

Budget

by Tencent

As of November 19, 2026, Hy4 flash by Tencent costs $0.07 per 1M input tokens and $0.28 per 1M output tokens, with a 1M-token context window.

Hy4 flash is the cheapest model Tencent sells and the one that finally delivers the open weights the full Hy4 promised and did not ship. Tencent launched it on November 17, 2026, and the international Tencent Cloud TokenHub list carries it at $0.070 per million input tokens, $0.003 on a cache hit, and $0.281 output against $0.834, $0.042, and $2.501 for Hy4, roughly a twelfth of the flagship rate on input and a ninth on output, with a 1,048,576 token context window. TokenHub halves those rates through November 30, 2026 and the model is free inside WorkBuddy over the same window. The architecture is where the cost comes from, and the two published counts differ: Tencent's announcement gives 224 billion total parameters including 64 billion Engram parameters with 12 billion active per token, while the Hugging Face model card counts 220 billion total, 155.5 billion excluding Engram, ViT, and DFly2, and splits the active budget into 11.3 billion text and 0.43 billion vision. Input is text and images, output is text. Weights went up on Hugging Face, ModelScope, and GitCode as Hy4-flash, Hy4-flash-FP8, and Hy4-flash-base on November 18, a day after a press release that still described open-sourcing as a plan, under the Tencent Hy License rather than the plain Apache 2.0 that the August Hy4 preview carried, so the terms need reading before any commercial deployment. Availability spans TokenHub, WorkBuddy, Yuanbao, WeCom, QQ, ima, and Marvis, with OpenRouter listed as coming soon. Every performance figure is Tencent's own: it reports Hy4 flash slightly ahead of the larger Hy4 Preview on an internal WorkIndex benchmark at about one thirteenth of the per-task cost, one twentieth of Hy4, and one eighth of DeepSeek-V4.1-Flash-1017, decoding above 150 tokens per second, a KV cache one twenty-fourth the size of Hy3's and one sixth of Hy4's, reinforcement learning that cuts average output tokens by half, and a blind test with 85 internal WorkBuddy users across 259 tasks where satisfaction landed close to Claude Haiku 5.5 at the lowest cost and shortest completion time of the models tried. No independent evaluator has scored it.

Input Price

$0.07

per 1M tokens

Output Price

$0.28

per 1M tokens

Context Window

1.0M

tokens

Released

2026-11

Open source

Capabilities

textvisioncodetool-usereasoning

Key Strengths

  • ✓$0.07 / $0.281 per 1M tokens with $0.003 cache hits, about a twelfth of Hy4 on input
  • ✓50 percent off on TokenHub and free in WorkBuddy through November 30, 2026
  • ✓Open weights on Hugging Face, ModelScope, and GitCode under the Tencent Hy License
  • ✓224B total parameters with only 12B active per token, vendor-stated
  • ✓1,048,576 token context window with text and image input
  • ✓Decoding above 150 tokens per second, vendor-reported

Best For

  • ▸High-volume everyday workflow automation at the lowest Hy rate
  • ▸Chart and document reading where the Hy4 preview was text only
  • ▸Self-hosted deployment, including the FP8 and base checkpoints
  • ▸Chinese-language productivity work through WorkBuddy, Yuanbao, WeCom, and QQ

Pricing Details

Input tokens

$0.07

per 1M tokens

Output tokens

$0.28

per 1M tokens

Estimated cost per 1K requests

$0.21

~1K input + ~500 output tokens avg

Prices are subject to change. Check the official documentation for current pricing. See the cost calculator for detailed estimates.

Open Source Model

Hy4 flash is free to download and self-host under the Tencent Hy License. Hosted API pricing varies by provider (e.g., Together, Fireworks, Groq). See our open source LLM guide for deployment options.

Quick answers

How much does Hy4 flash cost?
As of November 19, 2026, Hy4 flash costs $0.07 per 1M input tokens and $0.28 per 1M output tokens on the Tencent API.
What is the context window of Hy4 flash?
Hy4 flash has a 1M-token context window (1,048,576 tokens).

Related Models

View DocumentationCompare ModelsCost CalculatorFull Pricing Guide