Skip to main content
Watch: I Built an AI App Without Leaving Databricks

Tencent - Open weights

Hy3 pricing and specs

Hy3 is an open-weights model from Tencent that costs $0.132 per 1M input tokens and $0.528 per 1M output tokens on the first-party API, with a 256K-token context window.

Last verified from tencentcloud.com. All sources

Input / 1M
$0.132
Output / 1M
$0.528
Context
256K tokens
Weights
Open

Pricing per 1M tokens

ItemPrice / 1M tokens
Input$0.132
Output$0.528
Cache read$0.033

Note: Sold through Tencent Cloud TokenHub (international). The 'Cache' column is recorded as cache read. No cache-write, batch or peak/off-peak price is listed. The price is the same in both regional language-model tables (Singapore/Guangzhou tabs). The model list states the context window as 256k (max input 192k, max output 128k), recorded here in decimal as 256000/128000; the Hugging Face card says 'Context Length 256K'. 295B total / 21B active parameter MoE. Release date is the publish date of Tencent's official Hy3 launch post (2026-07-06). Hy3 preview came earlier, on 23 April 2026, and is a separate model. Vercel AI Gateway shows a different price for Hy3 (0.14/0.58, varies by provider); the first-party TokenHub price is recorded here.

Specs

Context window
256,000 tokens
Max output
128,000 tokens
Release date
Provider
Tencent
API model id
hy3
Vercel AI Gateway id
tencent/hy3
Weights
Open weights
License
Apache-2.0
Weights download
huggingface.co

What Hy3 costs in practice

Worked from the standard list rates above, with the math shown so you can swap in your own token counts.

One agent session

1M input tokens and 100K output tokens, no caching.

  • Input1M x $0.132 = $0.132
  • Output0.1M x $0.528 = $0.0528

$0.1848per session

A chat app

10,000 requests a day at 2K input and 500 output tokens each.

  • Input20M x $0.132 = $2.64
  • Output5M x $0.528 = $2.64

$5.28per day

$158.40 per 30-day month

The same session at a 90% cache hit

90% of the input is read from the prompt cache. The one-time cache write premium is not included.

  • Cached input0.9M x $0.033 = $0.0297
  • Uncached input0.1M x $0.132 = $0.0132
  • Output0.1M x $0.528 = $0.0528

$0.0957per session

Compare with

ModelInput / 1MOutput / 1M
Hy3 (this page)$0.132$0.528
Claude Haiku 5.5Anthropic$0.10$0.50
GLM-5.3-FlashZ.ai$0.15$0.50
GPT-6 LunaOpenAI$0.10$0.50
Qwen-MT-FlashAlibaba Cloud (Qwen)$0.16$0.49
Hy-MT2-PlusTencent$0.074$0.295
Hy-MT2-ProTencent$0.074$0.295
Hy-MT2-LiteTencent$0.044$0.177
Hy4 previewTencent$0.834$2.501

Related reading

FAQ

How much does Hy3 cost per million tokens?

Hy3 costs $0.132 per 1M input tokens and $0.528 per 1M output tokens on Tencent's first-party API, per tencentcloud.com on October 17, 2026. Per 1M tokens, cache reads cost $0.033.

What is Hy3's context window?

Hy3 has a 256,000-token context window and returns up to 128,000 output tokens per response.

Are Hy3's weights open?

Yes. Hy3's weights are published at https://huggingface.co/tencent/Hy3 under the Apache-2.0.

When was Hy3 released?

Tencent released Hy3 on July 6, 2026.

Sources

Last verified from tencentcloud.com.

Newsletter

Get model price changes as they ship

One email per week. New model releases, price cuts and what they mean for your bill.