Hugging Face's logo
Activity Feed

AI & ML interests

Quantization-aware training (QAT), low-bit LLMs (INT4, INT3, INT2, NVFP4), quantization-aware distillation

Recent Activity
vincentcounathe  updated a Space 22 days ago
QUASAR-QAT/README
vincentcounathe  published a Space 22 days ago
QUASAR-QAT/README
vincentcounathe  updated a model 22 days ago
QUASAR-QAT/gemma-4-E4B-it-QUASAR-W4A16-G64
View all activity

Papers

QUASAR: Lowering the Loss Floor of Quantization-Aware Training with Loss-Aware Reconstruction

View all Papers

QUASAR-QAT 's collections 4

Qwen3.5-4B & Qwen3.8-27B — QUASAR 4-bit QAT
Full-decoder 4-bit Qwen with QUASAR: 27B leads the compared NVFP4 builds on reasoning; 4B ships W4A16, Blackwell W4A4, and native Q4_0 GGUF.
Muse Glimmer 30B — QUASAR 4-bit QAT
Muse Glimmer 30B at 4 bits with QUASAR: W4A16 72% lower KL than Red Hat / 39% than NVIDIA · W4A4 20% lower than Red Hat · Q4_0 GGUF beats Meta Q4_K_M.
Gemma 4 E4B & 12B — QUASAR 4-bit QAT
Native 4-bit Gemma 4 E4B & 12B with QUASAR: lower KL to BF16 than Google's official QAT at both sizes, in Q4_0 GGUF (llama.cpp) and W4A16 (vLLM).
All QUASAR Models
All QUASAR checkpoints in one place: 4-bit QAT for Qwen, Gemma and Muse — NVFP4 W4A16/W4A4 for vLLM, Q4_0 GGUF for llama.cpp / Ollama.
Qwen3.5-4B & Qwen3.8-27B — QUASAR 4-bit QAT
Full-decoder 4-bit Qwen with QUASAR: 27B leads the compared NVFP4 builds on reasoning; 4B ships W4A16, Blackwell W4A4, and native Q4_0 GGUF.
Gemma 4 E4B & 12B — QUASAR 4-bit QAT
Native 4-bit Gemma 4 E4B & 12B with QUASAR: lower KL to BF16 than Google's official QAT at both sizes, in Q4_0 GGUF (llama.cpp) and W4A16 (vLLM).
Muse Glimmer 30B — QUASAR 4-bit QAT
Muse Glimmer 30B at 4 bits with QUASAR: W4A16 72% lower KL than Red Hat / 39% than NVIDIA · W4A4 20% lower than Red Hat · Q4_0 GGUF beats Meta Q4_K_M.
All QUASAR Models
All QUASAR checkpoints in one place: 4-bit QAT for Qwen, Gemma and Muse — NVFP4 W4A16/W4A4 for vLLM, Q4_0 GGUF for llama.cpp / Ollama.