Hy3 GGUF size and VRAM requirements

License: apache-2.0 ⬇ 10,937 ❤ 5
Parameters298.79B
Context262,144

bartowski/Hy3-GGUF is a very large language model with 298.79 billion parameters, built on the hy_v3 architecture. It is released under the apache-2.0 license and has been downloaded 10,937 times.

To run bartowski/Hy3-GGUF locally at a 4,096-token context, its quantized versions need between 9.7 GB (GGUF, lowest quality) and 304.98 GB (Q8_0, highest quality) of memory, weights plus KV cache and a system margin included.

For most users the best balance is GGUF, needing about 9.7 GB. That means bartowski/Hy3-GGUF fits entirely in the VRAM of a 10 GB GPU or larger, running fully on the GPU.

Available GGUF quantizations for bartowski/Hy3-GGUF include GGUF, IQ1_S, IQ1_M, IQ2_XXS, IQ2_XS, IQ2_S, IQ2_M, Q2_K, Q2_K_L, IQ3_XXS, Q3_K_S, IQ3_XS, Q3_K_M, Q3_K_L, Q3_K_XL, IQ3_M, IQ4_XS, IQ4_NL, Q4_0, Q4_K_S, Q4_K_M, Q4_K_L, Q4_1, Q5_K_S, Q5_K_M, Q6_K, Q8_0. The model supports a native context length of up to 262,144 tokens; a longer context grows the KV cache and the memory needed.

→ Guide: How much VRAM do you need?

GGUF file size and memory by quantization

Compare real GGUF weight sizes, estimated KV cache and total memory for Q4, Q5, Q8 and every quantization published in this repository.

Quant.Bits QualityWeights KVTotal Speed~Verdict
GGUF 0.02 Very low 0.56 GB 8.34 GB 9.7 GB 89.2 t/s Offload
IQ1_S 1.71 Very low 59.65 GB 8.34 GB 68.79 GB Insufficient
IQ1_M 1.9 Very low 66.22 GB 8.34 GB 75.36 GB Insufficient
IQ2_XXS 2.2 Very low 76.47 GB 8.34 GB 85.62 GB Insufficient
IQ2_XS 2.44 Very low 84.81 GB 8.34 GB 93.95 GB Insufficient
IQ2_S 2.48 Very low 86.43 GB 8.34 GB 95.57 GB Insufficient
IQ2_M 2.73 Low 95.08 GB 8.34 GB 104.22 GB Insufficient
Q2_K 2.86 Low 99.36 GB 8.34 GB 108.5 GB Insufficient
Q2_K_L 2.87 Low 99.81 GB 8.34 GB 108.95 GB Insufficient
IQ3_XXS 3.38 Fair 117.43 GB 8.34 GB 126.57 GB Insufficient
Q3_K_S 3.51 Fair 122.26 GB 8.34 GB 131.4 GB Insufficient
IQ3_XS 3.68 Fair 128.0 GB 8.34 GB 137.14 GB Insufficient
Q3_K_M 3.68 Fair 128.1 GB 8.34 GB 137.24 GB Insufficient
Q3_K_L 3.83 Fair 133.25 GB 8.34 GB 142.39 GB Insufficient
Q3_K_XL 3.84 Fair 133.65 GB 8.34 GB 142.8 GB Insufficient
IQ3_M 3.85 Fair 133.8 GB 8.34 GB 142.94 GB Insufficient
IQ4_XS 4.3 Good 149.73 GB 8.34 GB 158.87 GB Insufficient
IQ4_NL 4.55 Good 158.14 GB 8.34 GB 167.28 GB Insufficient
Q4_0 4.56 Good 158.62 GB 8.34 GB 167.76 GB Insufficient
Q4_K_S 4.7 Good 163.41 GB 8.34 GB 172.55 GB Insufficient
Q4_K_M 4.88 Good 169.65 GB 8.34 GB 178.79 GB Insufficient
Q4_K_L 4.89 Good 169.99 GB 8.34 GB 179.13 GB Insufficient
Q4_1 5.03 Very good 174.9 GB 8.34 GB 184.04 GB Insufficient
Q5_K_S 5.52 Very good 191.88 GB 8.34 GB 201.03 GB Insufficient
Q5_K_M 5.7 Very good 198.22 GB 8.34 GB 207.36 GB Insufficient
Q6_K 6.89 Excellent 239.58 GB 8.34 GB 248.72 GB Insufficient
Q8_0 8.51 Excellent 295.84 GB 8.34 GB 304.98 GB Insufficient

KV cache estimated (architecture unavailable). Speed is a rough estimate bounded by memory bandwidth.

Frequently asked questions

What kind of model is bartowski/Hy3-GGUF?

bartowski/Hy3-GGUF is a language model with 298.79 billion parameters, based on the hy_v3 architecture. It is released under the apache-2.0 license and distributed as GGUF files for local inference.

How much VRAM do you need to run bartowski/Hy3-GGUF?

You need about 9.7 GB of VRAM to run bartowski/Hy3-GGUF entirely on the GPU using the GGUF quantization (at a 4,096-token context). Smaller quantizations lower the requirement at the cost of quality.

Can I run bartowski/Hy3-GGUF on an 8 GB GPU?

Partially. bartowski/Hy3-GGUF only fits on an 8 GB GPU by offloading part of it to system RAM (with GGUF), which runs but is slower.

Can I run bartowski/Hy3-GGUF on a 16 GB GPU?

Yes. With 16 GB of VRAM you can run bartowski/Hy3-GGUF fully on the GPU using GGUF (about 9.7 GB).

Can I run bartowski/Hy3-GGUF on a 24 GB GPU?

Yes. With 24 GB of VRAM you can run bartowski/Hy3-GGUF fully on the GPU using GGUF (about 9.7 GB).

What context length does bartowski/Hy3-GGUF support?

bartowski/Hy3-GGUF supports a native context length of up to 262,144 tokens. A longer context grows the KV cache, so it increases the memory needed to run the model.

What is the best quantization for bartowski/Hy3-GGUF?

For bartowski/Hy3-GGUF, a strong default is Q4_K_M, which needs about 178.79 GB and keeps most of the quality while roughly halving the memory versus 8-bit. With VRAM to spare, Q5_K_M or Q6_K add a little more quality; if you are tight on memory, a smaller quantization still runs. Pick the highest quantization that fits your VRAM.