GraySoft
Projects Models Compare Cloud benchmarks FAQ Download guIDE →
Model Intelligence Sheet

tarruda/Hy3-GGUF overview

Hy3 GGUF My custom IQ3 XXS GGUF quantization of tencent/Hy3 https://huggingface.co/tencent/Hy3 . Quantization recipe | Recipe | Quant size | BPW | Default type…

gguftext-generationbase_model:tencent/Hy3base_model:quantized:tencent/Hy3license:apache-2.0endpoints_compatibleregion:usimatrixconversational

Runs locally from ~4.8 MB disk (4 GB VRAM class GPUs with llama.cpp / guIDE).

Downloads
0
Likes
1
Pipeline
text-generation
Author

Repository Files & Downloads

6 GGUF files detected
Direct downloads for local inference
FileTypeQuantizationSizeLink
Hy3-MTP-Q8_0.ggufGGUFQ8_04.70 GBDownload
IQ3_XXS/Hy3-IQ3_XXS-00001-of-00004.ggufGGUFIQ3_XXS4.8 MBDownload
IQ3_XXS/Hy3-IQ3_XXS-00002-of-00004.ggufGGUFIQ3_XXS46.54 GBDownload
IQ3_XXS/Hy3-IQ3_XXS-00003-of-00004.ggufGGUFIQ3_XXS46.48 GBDownload
IQ3_XXS/Hy3-IQ3_XXS-00004-of-00004.ggufGGUFIQ3_XXS15.91 GBDownload
imatrix.ggufGGUFGGUF573.7 MBDownload

Model Details

Model IDtarruda/Hy3-GGUF
Authortarruda
Pipelinetext-generation
Licenseapache-2.0
Base modeltencent/Hy3
Last modified2026-07-14T12:44:15.000Z

Model README

---

quantized_by: tarruda

pipeline_tag: text-generation

base_model: tencent/Hy3

base_model_relation: quantized

license: apache-2.0

---

Hy3 GGUF

My custom IQ3_XXS GGUF quantization of

tencent/Hy3.

Quantization recipe

| Recipe | Quant size | BPW | Default type | Tensor-specific overrides |

| --- | ---: | ---: | --- | --- |

| IQ3_XXS | 111550.48 MiB | 3.17 | Q6_K | ffn_down_exps=iq3_xxs, ffn_gate_exps=iq3_xxs, ffn_up_exps=iq3_xxs |

Usage

#!/bin/sh -e

model=./IQ3_XXS/Hy3-IQ3_XXS-00001-of-00004.gguf

ctx=80000
parallel=1

ctx_size=$((ctx * parallel))

llama-server --no-mmap --no-warmup \
  --model "$model" --ctx-size "$ctx_size" -np "$parallel" \
  -ctk q8_0 -ctv q8_0 \
  --temp 0.9 --top-p 1.0 \
  --chat-template-kwargs '{"reasoning_effort":"high"}' \
  --cache-ram 0 -ctxcp 2 \
  --host 0.0.0.0

Run tarruda/Hy3-GGUF with guIDE

Download guIDE — the AI-native code editor with local LLM inference and 69 built-in tools.

Download guIDE → · Browse 524k+ models · Compare models

Source: Hugging Face · Compare models