Hy3

by Tencent · tencent/hy3

Hy3 is a 295B-parameter Mixture-of-Experts model from Tencent (21B active, 192 experts with top-8 routing) built for reasoning, agentic workflows, and real-world production use. It supports a configurable reasoning effort:...

Input / 1M tokens
$0.0825
Output / 1M tokens
$0.33
Context
262K

Specs

Input price$0.0825 / 1M tokens
Output price$0.33 / 1M tokens
Cached input$0.0206 / 1M tokens
Context window262K tokens
Max output128K tokens
Readstext
Writestext
Tool callingYes
ReasoningYes
ListedJul 6, 2026

List prices in USD, updated Oct 30, 2026. Providers may charge differently.

Providers

5 providers serve this model, as of Oct 30, 2026, 18:12 UTC.

Providers serving this model, sorted by price
ProviderQuantizationContextMax outputInput / 1MOutput / 1MCache read / 1MUptime (30 min)Latency (median)Throughput (median)SupportsStatus
TencentCheapestfp8262K128K$0.0825$0.33$0.0206100%2.72 s65 tok/sNormal
GMICloudbf16262K236K$0.14$0.58$0.035100%2.27 s26 tok/sNormal
Novita—262K236K$0.14$0.58$0.035100%1.97 s80 tok/sNormal
Phala—262K236K$0.15$0.64$0.04100%1.87 s61 tok/sNormal
AtlasCloudFastestfp8262K131K$0.20$0.80$0.05100%1.47 s106 tok/sNormal

"Cheapest" is the lowest input plus output price; "Fastest" the highest median throughput. Columns empty for every provider are not shown.

Usage

#15
by tokens, Oct 29, 2026
460B
tokens that day · 7 days 0%

Data as of 2026-10-30T18:12:03.879Z.All rankings