Hunyuan Hy4 API Pricing
Hy4 is the generally available row of Tencent's largest model, the 770B total / 49B active MoE first shipped as Hy4 Preview in August 2026. Tencent Cloud International lists it at $0.834/M input, $2.50/M output and $0.042/M cache hit - the same list as the preview - and through December 31, 2026 bills it at half that: $0.417 / $1.25 / $0.021. Unlike the preview, its model-list entry includes image understanding.
Hunyuan Hy4 API cost calculator.
Calculator pre-loaded with Hy4's list price - what you pay from January 1, 2027. Through December 31, 2026 Tencent bills Hy4 at 50% of list, so halve every result for the bill you get today. The cache slider matters: a cache hit is 95% below fresh input.
Hunyuan Hy4 cost on real workloads.
Codebase-scale migration
Reading 100-page contracts
Ticket triage
Research planning turn
Counts use a chars-per-token calibration measured on the vendor's own published tokenizer (tencent/Hunyuan-A13B-Instruct, 2026-06-10). English prose is typically within a few percent; code and non-Latin scripts tokenize heavier. For billing-exact counts use the vendor's count-tokens API.
| Model | Input /M | Output /M | Effective blended | Context | Best for |
|---|---|---|---|---|---|
| Hunyuan Hy4 Current | $0.834 cache $0.042 | $2.50 | $0.37 agentic 92/8 | 1M | Tencent's largest model, GA, with image input |
| Hunyuan Hy4 Preview | $0.834 cache $0.042 | $2.50 | $0.37 same | 1M | Same weights and list price, text-only preview row |
| Hunyuan Hy3 | $0.132 cache $0.033 | $0.528 | $0.089 −4.2× | 256K | Production traffic at a lower list price |
| GLM-5.3 | $1.40 cache $0.26 | $4.40 | $0.78 +2.1× | 1M | Coding-first frontier alternative |
| Kimi K3.1 Pro | $2.00 cache $0.20 | $10.00 | $1.28 +3.5× | 1M | Long-horizon agentic reasoning |
| Qwen3.8 Max | $2.00 | $6.00 | $2.32 +6.3× | 1M | Alibaba's flagship on the International list |
Hunyuan Hy4 API pricing FAQ.
What Hy4 costs today and from January, how it differs from Hy4 Preview, and what Tencent has and has not published about it.
Q · 01 What does Hy4 cost? +
$0.834/M input, $2.50/M output and $0.042/M on a cache hit. A notice on the same price list says Hy4 inference is settled at 50% of the list price from the notice date until 23:59:59 on December 31, 2026 (Beijing time), so the bill today is $0.417 / $1.25 / $0.021. From January 1, 2027 the list price applies again. This page stores the list price, so its calculator shows what you pay once the discount ends.Q · 02 How is Hy4 different from Hy4 Preview? +
hy4, and Tencent's model list adds image understanding to its capabilities, which the Hy4 Preview entry doesn't list. Context (1M, with 960k input and 64k output) is the same on both. Tencent's discount notice names Hy4 and doesn't say whether it also covers the preview row.Q · 03 Did Tencent announce Hy4? +
hy4 as a separate row. The release date on this page is the day we first verified the row, not a launch date from Tencent.Q · 04 How does it compare with Hy3? +
$0.132/$0.528 with a 256k context, against Hy4's $0.834/$2.50 and 1M. Cache hits close part of the gap - Hy4's cache rate is $0.042 against Hy3's $0.033 - so on cache-heavy agentic traffic the blended difference is smaller than the input rates suggest. Through December 31 the 50% discount narrows it further.