NVFP4 - a AutomatosX Collection

AutomatosX 's Collections

AXQ Reference Storage

NVFP4

updated 7 days ago

NVFP4 W4A16/W4A4 development packs; CPU-reference exports are identified per card. CUDA runtime evidence is model-specific.

AutomatosX/AX-Qwen3.8-27B-CUDA-AXQ-NVFP4-W4A16-MTP

Image-Text-to-Text • 28B • Updated 20 days ago • 104

Note NVFP4 W4A16 development artifact. See manifest-bound runtime evidence; MTP and quality are not certified.
AutomatosX/AX-Qwen3.5-9B-CUDA-AXQ-NVFP4-W4A16-MTP

Image-Text-to-Text • 10B • Updated 20 days ago • 46

Note NVFP4 W4A16 development artifact. See manifest-bound runtime evidence; MTP and quality are not certified.
AutomatosX/AX-Muse-Glimmer-30B-CUDA-AXQ-NVFP4-W4A16

Image-Text-to-Text • 30B • Updated 20 days ago • 63

Note NVFP4 W4A16 development artifact. See manifest-bound runtime evidence; MTP and quality are not certified.
AutomatosX/AX-Qwen3-Embedding-0.6B-CUDA-AXQ-NVFP4-W4A4

Sentence Similarity • 0.6B • Updated 28 days ago • 48

Note Native AXQuant CUDA NVFP4 W4A4 retrieval pack. Two-GPU development checks; not certified.
AutomatosX/AX-Qwen3-Embedding-4B-CUDA-AXQ-NVFP4-W4A4

Sentence Similarity • 4B • Updated 28 days ago • 45

Note Native AXQuant CUDA NVFP4 W4A4 retrieval pack. Two-GPU development checks; not certified.
AutomatosX/AX-Qwen3-Embedding-8B-CUDA-AXQ-NVFP4-W4A4

Sentence Similarity • 8B • Updated 28 days ago • 46

Note Native AXQuant CUDA NVFP4 W4A4 retrieval pack. Two-GPU development checks; not certified.
AutomatosX/AX-Nemotron-3-Embed-1B-CUDA-AXQ-NVFP4-W4A4

Sentence Similarity • 1B • Updated 28 days ago • 40

Note Native AXQuant CUDA NVFP4 W4A4 retrieval pack. Two-GPU development checks; not certified.
AutomatosX/AX-Nemotron-3-Embed-8B-CUDA-AXQ-NVFP4-W4A4

Sentence Similarity • 8B • Updated 28 days ago • 42

Note Native AXQuant CUDA NVFP4 W4A4 retrieval pack. Two-GPU development checks; not certified.
AutomatosX/AX-EmbeddingGemma-2-CUDA-AXQ-NVFP4-W4A16

Feature Extraction • 0.7B • Updated 23 days ago • 53

Note Native AXQuant CUDA NVFP4 W4A16 EmbeddingGemma 2 pack. Activations stay BF16. Not certified.
AutomatosX/AX-DeepSeek-OCR-2-CUDA-AXQ-NVFP4-W4A4

Image-Text-to-Text • 3B • Updated 28 days ago • 56

Note Native AXQuant CUDA NVFP4 W4A4. Two-GPU development smoke; no MTP or quality certification.
AutomatosX/AX-Qwen3-VL-4B-Instruct-CUDA-AXQ-NVFP4-W4A4

Image-Text-to-Text • 4B • Updated 28 days ago • 43

Note Native AXQuant CUDA NVFP4 W4A4. Two-GPU development smoke; no MTP or quality certification.
AutomatosX/AX-Qwen3-VL-8B-Instruct-CUDA-AXQ-NVFP4-W4A4

Image-Text-to-Text • 9B • Updated 28 days ago • 41

Note Native AXQuant CUDA NVFP4 W4A4. Two-GPU development smoke; no MTP or quality certification.
AutomatosX/AX-Clef-Flash-9B-CUDA-AXQ-NVFP4-W4A16

Text Generation • 9B • Updated 23 days ago • 35

Note Native AXQuant CUDA NVFP4 W4A16 Clef pack. Joint schema head stays BF16. Not certified.
AutomatosX/AX-Clef-27B-CUDA-AXQ-NVFP4-W4A16

Text Generation • 27B • Updated 23 days ago • 43

Note Native AXQuant CUDA NVFP4 W4A16 Clef pack. Joint schema head stays BF16. Not certified.
AutomatosX/AX-Swift-1.5-Qwen3.8-27B-CUDA-AXQ-NVFP4-W4A16-MTP

Image-Text-to-Text • 28B • Updated 21 days ago • 99

Note AXQ Swift-1.5 Qwen3.8-27B finetune. Vision and MTP stay BF16. Swift Open License v1.0; commercial use above US$1M revenue is not licensed. Not certified.
AutomatosX/AX-Hy-MT2-30B-A3B-CUDA-AXQ-NVFP4-W4A16

Translation • 30B • Updated 21 days ago • 63

Note AXQ Hy-MT2 pack. Norms and expert_bias stay BF16. Translation quality was not measured. Not certified.
AutomatosX/AX-Clef-Omni-30B-A3B-CUDA-AXQ-NVFP4-W4A16

Text Generation • 35B • Updated 23 days ago • 38

Note Native AXQuant CUDA NVFP4 W4A16 Clef-Omni pack. Vision, the audio encoder, code2wav, and the joint head stay BF16. Not certified.
AutomatosX/AX-Qwen3.5-0.8B-decision-CUDA-AXQ-NVFP4-W4A16

Text Generation • 0.9B • Updated 23 days ago • 36

Note Native AXQuant CUDA NVFP4 W4A16. Development artifact; not certified.
AutomatosX/AX-MiniCPM-V-4.7-35B-A3B-CUDA-AXQ-NVFP4-W4A16

Image-Text-to-Text • 35B • Updated 13 days ago • 36

Note NVFP4 W4A16 storage encoded by the CPU reference backend; fused 3D experts and vision stay BF16. CUDA runtime and quality are unverified.
AutomatosX/AX-DeepSeek-OCR-2-CUDA-AXQ-NVFP4-W4A16

Image-Text-to-Text • 3B • Updated 28 days ago • 58

Note Native AXQuant CUDA NVFP4 W4A16. Development artifact; not certified.
AutomatosX/AX-Unlimited-OCR-3B-MoE-CUDA-AXQ-NVFP4-W4A16

Image-Text-to-Text • 3B • Updated 28 days ago • 54

Note Native AXQuant CUDA NVFP4 W4A16. Development artifact; not certified.
AutomatosX/AX-Nemotron-3-Super-120B-A12B-CUDA-AXQ-NVFP4-MTP

Text Generation • 124B • Updated 28 days ago • 438

Note Native AXQuant CUDA NVFP4 W4A16. Development artifact; not certified.
AutomatosX/AX-Nemotron-3.5-Lightning-30B-A3B-CUDA-AXQ-NVFP4-MTP

Text Generation • 33B • Updated 28 days ago • 491

Note Native AXQuant CUDA NVFP4 W4A16. Development artifact; not certified.
AutomatosX/AX-Unlimited-OCR-2-CUDA-AXQ-NVFP4-W4A16

3B • Updated 7 days ago • 53 • 1

Note NVFP4 W4A16 storage encoded by the CPU reference backend; fused 3D experts and vision stay BF16. CUDA runtime and quality are unverified.