Aleph Alpha Kolibri 1, 78B-A3.5B MoE at 262k ctx - serving needs the aleph-alpha-inference vLLM plugin as well as glq.
Christian N PRO
xv0y5ncu
AI & ML interests
GLQ E8 Quantization, reasoning AI software. MLops. Huggingface Transformers, vLLM & SGLang. Langchain, RAG. Open source software. Kubernetes
Making LLMs run better on more limited devices such as gaming GPUs
Recent Activity
updated a model 2 days ago
xv0y5ncu/Kolibri-1-GLQ-trellis-3inst-3bpw upvoted an article 2 days ago
Extending Nemotron 3.5 Super 120B VL with Agentic Web Development Capabilities updated a collection 2 days ago
Kolibri 1 - GLQ quantizedOrganizations
None yet
All public Gemma-4 GLQ checkpoints: 12B, 26B-A4B (MoE), 31B, E4B - 3 to 8 bpw.
-
xv0y5ncu/gemma-4-26B-A4B-it-GLQ-4bpw
Image-Text-to-Text • 7B • Updated • 32 -
xv0y5ncu/gemma-4-26B-A4B-it-GLQ-3bpw
Image-Text-to-Text • 7B • Updated • 32 -
xv0y5ncu/gemma-4-26B-A4B-it-GLQ-5bpw
Image-Text-to-Text • 11B • Updated • 27 -
xv0y5ncu/gemma-4-26B-A4B-it-GLQ-6bpw
Image-Text-to-Text • 11B • Updated • 28
SmolLM2 135M/360M and SmolLM3-3B - the small end: CI, demos, and cheap codebook A/Bs.
-
xv0y5ncu/SmolLM3-3B-trellis-3inst-4bpw-kernel
Text Generation • 1.0B • Updated • 13.6k -
xv0y5ncu/SmolLM3-3B-GLQ-block-diagonal-3.5bpw
Text Generation • 0.9B • Updated • 243 -
xv0y5ncu/SmolLM3-3B-GLQ-3.5bpw
Text Generation • 1B • Updated • 358 -
xv0y5ncu/SmolLM3-3B-GLQ-6bpw
Text Generation • 1B • Updated • 216
One curated pick per model family - current codebooks, true-to-label footprints. pip install glq
-
xv0y5ncu/SmolLM3-3B-trellis-3inst-4bpw-kernel
Text Generation • 1.0B • Updated • 13.6k -
xv0y5ncu/gemma-4-26B-A4B-it-GLQ-trellis-3inst-4bpw
7B • Updated • 190 -
xv0y5ncu/gemma-4-26B-A4B-it-GLQ-trellis-3inst-3bpw
6B • Updated • 35 -
xv0y5ncu/Gemma-4-31B-it-GLQ-5.0bpw-mix3-8
Text Generation • 12B • Updated • 427
One base model across many bit-widths - read the quality/footprint curve, not a single point.
Aleph Alpha Kolibri 1, 78B-A3.5B MoE at 262k ctx - serving needs the aleph-alpha-inference vLLM plugin as well as glq.
One curated pick per model family - current codebooks, true-to-label footprints. pip install glq
-
xv0y5ncu/SmolLM3-3B-trellis-3inst-4bpw-kernel
Text Generation • 1.0B • Updated • 13.6k -
xv0y5ncu/gemma-4-26B-A4B-it-GLQ-trellis-3inst-4bpw
7B • Updated • 190 -
xv0y5ncu/gemma-4-26B-A4B-it-GLQ-trellis-3inst-3bpw
6B • Updated • 35 -
xv0y5ncu/Gemma-4-31B-it-GLQ-5.0bpw-mix3-8
Text Generation • 12B • Updated • 427
All public Gemma-4 GLQ checkpoints: 12B, 26B-A4B (MoE), 31B, E4B - 3 to 8 bpw.
-
xv0y5ncu/gemma-4-26B-A4B-it-GLQ-4bpw
Image-Text-to-Text • 7B • Updated • 32 -
xv0y5ncu/gemma-4-26B-A4B-it-GLQ-3bpw
Image-Text-to-Text • 7B • Updated • 32 -
xv0y5ncu/gemma-4-26B-A4B-it-GLQ-5bpw
Image-Text-to-Text • 11B • Updated • 27 -
xv0y5ncu/gemma-4-26B-A4B-it-GLQ-6bpw
Image-Text-to-Text • 11B • Updated • 28
One base model across many bit-widths - read the quality/footprint curve, not a single point.
SmolLM2 135M/360M and SmolLM3-3B - the small end: CI, demos, and cheap codebook A/Bs.
-
xv0y5ncu/SmolLM3-3B-trellis-3inst-4bpw-kernel
Text Generation • 1.0B • Updated • 13.6k -
xv0y5ncu/SmolLM3-3B-GLQ-block-diagonal-3.5bpw
Text Generation • 0.9B • Updated • 243 -
xv0y5ncu/SmolLM3-3B-GLQ-3.5bpw
Text Generation • 1B • Updated • 358 -
xv0y5ncu/SmolLM3-3B-GLQ-6bpw
Text Generation • 1B • Updated • 216