Hy4

Model profile
Model	Size	Approximate size
Hy4Model source: provider1,048,576 tokens max native context	770Bactive 49B	462 GB4-bit quantization

Compatible hardware Estimated speeds, not benchmark results: calculated from memory bandwidth and model size. Real results can differ significantly because there is no precise formula for deriving LLM generation speed from hardware specifications alone.
Hardware	Memory	Bandwidth	vLLM	Estimated generation
DGX H200	1128 GBGPU HBM3e	4800 GB/s	yes	62 tok/s4-bit quantization
DGX Station	748 GBCoherent Memory	7100 GB/s	yes	88 tok/s4-bit quantization
ET900N G3	748 GBCoherent Memory	7100 GB/s	yes	88 tok/s4-bit quantization
DGX B200	1440 GBGPU HBM3e	8000 GB/s	yes	97 tok/s4-bit quantization
GB200 NVL72	13400 GBGPU HBM3e	8000 GB/s	yes	97 tok/s4-bit quantization
GB300 NVL72	20000 GBGPU HBM3e	8000 GB/s	yes	97 tok/s4-bit quantization

7 videos found Showing 1-7

Video Fahd Mirza 2026-09-17new

Tencent Shrank a 1.5TB AI Model to 214GB — Here's the Trick

Explains how Tencent used Sherry quantization to shrink a 770B model from 1.5TB to 214GB.

Hy4

Video xCreate 2026-09-05

I Ran Hy4 770B Locally for 48 Hours... Local AI is the Frontier! 🤯

Tests Hy4 Preview against GLM 5.3 using Inferencer App Q4-INF quantization on Mac Studio M3 Ultra 512 GiB via thinking mode performance, UI testing, censorship checks, complex code modification, game

Testing Hy4 M3 Ultra

Video xCreate 2026-09-03

Let's Try Hy4 Open Weight Model - is it Better than GLM 5.3? 😳

Tests Tencent's 700B+ MoE Hy4 using Workbuddy app to evaluate its maximum performance capabilities.

Testing Hy4

Video AI with Eric 2026-09-02

Tencent Shrunk Hy4 From 1.5 TB to 214 GB. I Tested It Locally

Tests Tencent Hy4 preview MIX-STQ1_0 GGUF locally against API versions via Slappis demo, Splunk CTF, and Estonia tasks, evaluating coding and agentic performance preservation.

Testing Hy4

Video WorldofAI 2026-09-01

Tencent HY4 IS SOLID! Best Open-Weight Model? (FULLY FREE)

Tests Tencent HY4 Preview on coding, 3D game dev, and agentic tasks, comparing it against Qwen 3.8, GLM 5.3, DeepSeek V4 Flash Vision, Kimi K3, Claude Opus 5, and GPT-5.6 using WOAI Bench.

Testing Hy4