RunsOnMy ModelsGPUs & MacsAPI pricesVRAM calculatorNew

Can I run Hy3 locally?

Hy3 by tencent has 299B parameters (21B active per token, 192 experts, 8 used per token). At 4-bit with an 8K context it needs about 173 GB of GPU memory. It fits on 6 of 65 GPUs and Macs we track.

173 GBQ4_K_M · 8K context
nonesmallest consumer GPU that fits
M3 Ultra 256GBsmallest Mac that fits
256Kmax context
GQAattention type

Memory needed by quant and context length

GB of GPU memory = weights + KV cache (FP16) + runtime overhead. Weights use real file sizes from bartowski/Hy3-GGUF where available (marked •), otherwise parameter count × bits per weight. How we calculate.

QuantQuality4K ctx8K ctx32K ctx128K ctx
2-bit (Q2_K) •rough101 GB102 GB110 GB140 GB
3-bit (Q3_K_M) •usable130 GB131 GB139 GB169 GB
4-bit (Q4_K_M) •good, most popular172 GB173 GB180 GB210 GB
5-bit (Q5_K_M) •very good200 GB201 GB209 GB239 GB
6-bit (Q6_K) •near lossless241 GB243 GB250 GB280 GB
8-bit (Q8_0) •effectively lossless298 GB299 GB306 GB336 GB
16-bit (F16/BF16)original weights558 GB560 GB567 GB597 GB

Which GPUs and Macs run Hy3?

Best quality quant that fits at 8K context. "Tight" = within 10% of usable memory. Speed is a ceiling from memory bandwidth (real speed is typically 50–80% of it).

HardwareMemoryFits?Best quantNeedsMax tok/s (est.)
RTX 3060 12GB12 GBNo———details
RTX 4060 Ti 16GB16 GBNo———details
RTX 309024 GBNo———details
RTX 409024 GBNo———details
RTX 509032 GBNo———details
RTX 5070 Ti16 GBNo———details
RX 7900 XTX24 GBNo———details
Mac M4 Pro 48GB48 GBNo———details
Mac M4 Max 128GB128 GBNo———details
Mac M3 Ultra 512GB512 GBFitsQ8_0299 GB37details
DGX Spark128 GBTightQ2_K102 GB36details
RTX 407012 GBNo———details
RTX 5060 Ti 16GB16 GBNo———details
RTX 40608 GBNo———details
Arc B58012 GBNo———details
RTX 3060 Ti8 GBNo———details
RTX 30708 GBNo———details
RTX 3080 10GB10 GBNo———details
RTX 3080 12GB12 GBNo———details
RTX 3090 Ti24 GBNo———details
RTX 4060 Ti 8GB8 GBNo———details
RTX 4070 Super12 GBNo———details
RTX 4070 Ti12 GBNo———details
RTX 4070 Ti Super16 GBNo———details
RTX 408016 GBNo———details
RTX 4080 Super16 GBNo———details
RTX 50608 GBNo———details
RTX 507012 GBNo———details
RTX 508016 GBNo———details
2x RTX 30902×24 GBNo———details
2x RTX 40902×24 GBNo———details
2x RTX 50902×32 GBNo———details
RTX A600048 GBNo———details
RTX 6000 Ada48 GBNo———details
RTX PRO 6000 Blackwell96 GBNo———details
L40S48 GBNo———details
A100 40GB40 GBNo———details
A100 80GB80 GBNo———details
H100 PCIe80 GBNo———details
H100 SXM80 GBNo———details
H200141 GBFitsQ2_K102 GB641details
RX 76008 GBNo———details
RX 7700 XT12 GBNo———details
RX 7800 XT16 GBNo———details
RX 7900 GRE16 GBNo———details
RX 7900 XT20 GBNo———details
RX 907016 GBNo———details
RX 9070 XT16 GBNo———details
Ryzen AI Max+ 395 128GB128 GBTightQ2_K102 GB34details
Arc A770 16GB16 GBNo———details
Mac M1 16GB16 GBNo———details
Mac M2 24GB24 GBNo———details
Mac M3 24GB24 GBNo———details
Mac M4 16GB16 GBNo———details
Mac M4 32GB32 GBNo———details
Mac M1 Pro 32GB32 GBNo———details
Mac M3 Pro 36GB36 GBNo———details
Mac M4 Pro 24GB24 GBNo———details
Mac M4 Pro 64GB64 GBNo———details
Mac M1 Max 64GB64 GBNo———details
Mac M2 Max 96GB96 GBNo———details
Mac M3 Max 128GB128 GBNo———details
Mac M4 Max 64GB64 GBNo———details
Mac M2 Ultra 192GB192 GBFitsQ2_K102 GB107details
Mac M3 Ultra 256GB256 GBFitsQ4_K_M173 GB64details

Prefer an API?

Tencent: Hy3 is on OpenRouter at $0.132 per 1M input tokens and $0.528 per 1M output tokens. Price details and history →

Questions

How much VRAM does Hy3 need?

About 173 GB at 4-bit (Q4_K_M) with an 8K context: 170 GB of weights (real GGUF file size), 2.5 GB of KV cache and 0.6 GB of runtime overhead. At 8-bit it needs about 299 GB.

What is the smallest GPU that runs Hy3?

No single consumer GPU in our table fits it at Q4_K_M with 8K context; the smallest option that does is the Mac M3 Ultra 256GB (256 GB).

Can I run Hy3 on a Mac?

Yes. At Q4_K_M with 8K context it fits on a Mac M3 Ultra 256GB or larger, using the default macOS GPU memory limit.

Model facts

Hugging Face repotencent/Hy3
Parameters298,786,155,776 (299B) · ~21B active (computed from config)
Layers / hidden size80 / 4096
AttentionGQA; 64 heads, 8 KV heads, head dim 128
KV cache per token (FP16)320.0 KiB
Max context262,144 tokens
Licenseapache-2.0
Released on HFJul 2, 2026
Downloads (30d) / likes987,079 / 993

Sources: Hugging Face model API · config.json (tencent/Hy3) · GGUF files · GPU specs from vendor pages (list). Data fetched Nov 17, 2026. Spot an error? Report it.