TensorFold/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-MLX-4bit
Text Generation • 32B • Updated • 4.39k • 5
MLX models for 64GB Macs
MLX models for 64GB Macs. See notes for measured versus estimated fit; context length and concurrent requests affect memory use.
Note Published M3 Studio measurement: 17.95 GB peak, 168.41 tokens/s median; three 256-token runs after warm-up. About 46 GB nominal headroom on 64GB; longer prompts need more memory. See the card for runtime and setup.
TensorFold/Qwen3.8-27B-MLX-4bit
Image-Text-to-Text • 27B • Updated • 7.79k • 1Note Published M3 Studio peak: 20.14 GB; median decode: 39.90 tokens/s. About 43 GB nominal headroom on 64GB. See the card for benchmark conditions, runtime and setup; context increases memory.
TensorFold/Muse-Glimmer-30B-MLX-4bit
Image-Text-to-Text • 30B • Updated • 1.24k • 2Note Published M3 Studio peak: 21.33 GB; median decode: 37.70 tokens/s. About 42 GB nominal headroom on 64GB. See the card for Muse runtime support and benchmark conditions.
TensorFold/Nex-N2.5-mini-MLX-4bit
Image-Text-to-Text • 35B • Updated • 141Note Estimated 64GB fit; tested on a 256GB Studio with oMLX 0.6.4, not a 64GB Mac. Peak request memory was not measured. Start with short context and one request; see the card for sampling settings and quality limitations.
TensorFold/Nex-N2.5-mini-MLX-6bit
Image-Text-to-Text • 35B • Updated • 121 • 1Note Estimated 64GB fit; tested on a 256GB Studio with oMLX 0.6.4, not a 64GB Mac. Peak request memory was not measured. Start with short context and one request; see the card for sampling settings and quality limitations.
TensorFold/Nex-N2.5-mini-MLX-8bit
Image-Text-to-Text • 35B • Updated • 135 • 1Note Estimated 64GB fit; tested on a 256GB Studio with oMLX 0.6.4, not a 64GB Mac. Peak request memory was not measured. Start with short context and one request; see the card for sampling settings and quality limitations.
TensorFold/Nex-N2.5-mini-MLX-oQ2
Image-Text-to-Text • 35B • Updated • 108 • 1Note Estimated 64GB fit; tested on a 256GB Studio with oMLX 0.6.4, not a 64GB Mac. Peak request memory was not measured. Start with short context and one request; see the card for sampling settings and quality limitations.
TensorFold/Nex-N2.5-mini-MLX-oQ3
Image-Text-to-Text • 35B • Updated • 68Note Estimated 64GB fit; tested on a 256GB Studio with oMLX 0.6.4, not a 64GB Mac. Peak request memory was not measured. Start with short context and one request; see the card for sampling settings and quality limitations.
TensorFold/Nex-N2.5-mini-MLX-oQ4
Image-Text-to-Text • 35B • Updated • 141Note Estimated 64GB fit; tested on a 256GB Studio with oMLX 0.6.4, not a 64GB Mac. Peak request memory was not measured. Start with short context and one request; see the card for sampling settings and quality limitations.
TensorFold/Nex-N2.5-mini-MLX-oQ6
Image-Text-to-Text • 35B • Updated • 111 • 1Note Estimated 64GB fit; tested on a 256GB Studio with oMLX 0.6.4, not a 64GB Mac. Peak request memory was not measured. Start with short context and one request; see the card for sampling settings and quality limitations.
TensorFold/Nex-N2.5-mini-MLX-oQ8
Image-Text-to-Text • 35B • Updated • 120 • 1Note Estimated 64GB fit; tested on a 256GB Studio with oMLX 0.6.4, not a 64GB Mac. Peak request memory was not measured. Start with short context and one request; see the card for sampling settings and quality limitations.