FITMYLLM · AUGUST 17, 2026 · OPEN METHODOLOGY · COMMUNITY BENCHMARKS
LIVE · UPDATED DAILY
295B
Parameters (21B active)
0
Quantization Options Select your GPU for speed estimates Context length: 4K 8K 16K 32K 64K 128K 256K
Quant Bits VRAM @ 16K Quality IQ2_XXS 2.38 92.0 GB
88.3 + 3.8 KV
low IQ2_M 2.93 112.3 GB
108.5 + 3.8 KV
low Q2_K 3.16 120.8 GB
117.0 + 3.8 KV
low IQ3_XXS 3.25 124.1 GB
120.3 + 3.8 KV
low IQ3_XS 3.5 133.3 GB
129.6 + 3.8 KV
low Q3_K_S 3.64 138.5 GB
134.7 + 3.8 KV
low IQ3_M 3.76 142.9 GB
139.1 + 3.8 KV
low Q3_K_M 4 151.7 GB
148.0 + 3.8 KV
low Q3_K_L 4.3 162.8 GB
159.1 + 3.8 KV
moderate IQ4_XS 4.46 168.7 GB
165.0 + 3.8 KV
moderate Q4_K_S 4.67 176.4 GB
172.7 + 3.8 KV
moderate Q4_K_M 4.89 184.6 GB
180.8 + 3.8 KV
good Q5_K_S 5.57 209.6 GB
205.9 + 3.8 KV
good Q5_K_M 5.7 214.4 GB
210.7 + 3.8 KV
good Q6_K 6.56 246.1 GB
242.4 + 3.8 KV
excellent Q8_0 8.5 317.7 GB
313.9 + 3.8 KV
lossless FP16 16 594.2 GB
590.5 + 3.8 KV
lossless
Select your GPU above to see speed estimates and compatibility for each quantization.
Too big for a single GPU — plan a multi-GPU deployment
Even the lightest quant needs ~92 GB. Size GPUs, replicas, TCO and scaling for a production setup. Open in Enterprise →
▸ READY TO RUN THIS? RENT BY THE HOUR
RENT A GPU AND RUN HUNYUAN HY3-PREVIEW NOW
Spin up an A100 / H100 / 4090 in ~60s. Pay by the second. Cancel anytime.
Community Ratings Chat Coding Reasoning Creative Vision Roleplay Agentic
Loading ratings...
Run this model IQ2_XXS — 88.3 GB VRAM IQ2_M — 108.5 GB VRAM Q2_K — 117.0 GB VRAM IQ3_XXS — 120.3 GB VRAM IQ3_XS — 129.6 GB VRAM Q3_K_S — 134.7 GB VRAM IQ3_M — 139.1 GB VRAM Q3_K_M — 148.0 GB VRAM Q3_K_L — 159.1 GB VRAM IQ4_XS — 165.0 GB VRAM Q4_K_S — 172.7 GB VRAM Q4_K_M — 180.8 GB VRAM Q5_K_S — 205.9 GB VRAM Q5_K_M — 210.7 GB VRAM Q6_K — 242.4 GB VRAM Q8_0 — 313.9 GB VRAM FP16 — 590.5 GB VRAM
Ollama llama.cpp vLLM LM Studio KoboldCpp Jan Docker
▸ Easiest way to get started · Beginners
DOCS ↗ curl -fsSL https://ollama.com/install.sh | shCOPY
$ ollama run hunyuan:295b-q4_K_MCOPY
Tag may need adjustment — check ollama.com/library/hunyuan for available tags.
▸ SETUP GUIDE >_
Auto-setup with fitmyllm CLI Detects your GPU, recommends the best model, downloads it, and starts chatting — zero config. Benchmarks your speed and contributes anonymous data to improve predictions.
Auto-detect GPU Live tok/s in chat Speed benchmarks 9 inference engines
GPUs that can run this model At Q4_K_M quantization. Sorted by minimum VRAM.
Find the best GPU for Hunyuan Hy3-preview
Build Hardware for Hunyuan Hy3-preview ▸ COLOPHON FITMYLLM · INDEPENDENT · DATA-DRIVEN
FITMYLLM · EST. 2025 · © 2026
RECOMMENDATIONS FROM PUBLISHED MATH, CORRECTED BY THE COMMUNITY — 30.
▸ SPEC SHEET
Hunyuan Hy3-preview — 295B MoE. ▸ SPECIFICATIONS
PARAMETERS 295B (21B active)
ARCHITECTURE Mixture of Experts
CONTEXT LENGTH 256K tokens
CAPABILITIES chat, coding, reasoning, multilingual, math, tool_use
RELEASE DATE 2026-04-23
PROVIDER Tencent
FAMILY hunyuan ▸ VRAM REQUIREMENTS
QUANT BPW VRAM QUALITY IQ2_XXS 2.38 88.3 GB 65% IQ2_M 2.93 108.5 GB 75% Q2_K 3.16 117.0 GB 78% IQ3_XXS 3.25 120.3 GB 82% IQ3_XS 3.5 129.6 GB 84% Q3_K_S 3.64 134.7 GB 85% IQ3_M 3.76 139.1 GB 86% Q3_K_M 4 148.0 GB 88% Q3_K_L 4.3 159.1 GB 90% IQ4_XS 4.46 165.0 GB 92% Q4_K_S 4.67 172.7 GB 93% Q4_K_M 4.89 180.8 GB 94% Q5_K_S 5.57 205.9 GB 96% Q5_K_M 5.7 210.7 GB 96% Q6_K 6.56 242.4 GB 97% Q8_0 8.5 313.9 GB 100% FP16 16 590.5 GB 100%
§ 01 BENCHMARK SCORES
MMLU-PRO 65.8
MATH 76.3
MBPP 78.7
GPQA Diamond 87.2
LiveCodeBench 34.9
HLE 30.0
aa_terminal_bench 54.4
SWE-bench 74.4
§ 03 COMPATIBLE GPUs
13 @ Q4_K_M Feedback