← CanIRunLocal · All hardware · All models
Alibaba · 125B parameters (6B active) · chat, reasoning, coding, vision.
Ollama weight files: Q4_K_M 120 GB (qwen3.8-flash-next:125b-a6b-q4_K_M); Q8_0 189 GB (qwen3.8-flash-next:125b-a6b-q8_0).
Yes on 15 machines in this list. The smallest is M3 Max 128GB: Q4_K_M, about 113.9 GB, roughly 30 tokens per second, via LM Studio. lms get mlx-community/Qwen3.8-Flash-Next-4bit
lms get mlx-community/Qwen3.8-Flash-Next-4bitollama run qwen3.8-flash-next:125b-a6b-q8_0ollama run qwen3.8-flash-next:125b-a6b-q8_0lms get mlx-community/Qwen3.8-Flash-Next-4bitllama-server -hf unsloth/Qwen3.8-Flash-Next-GGUF:Q8_0 -c 8192mlx_lm.generate --model mlx-community/Qwen3.8-Flash-Next-4bit --prompt "Hello"| Hardware | Runtime | Quant | Memory | Fit | Speed | Command |
|---|---|---|---|---|---|---|
| M3 Max 128GB | LM Studio | MLX 4bit | 113.9 GB | fits if the Mac GPU limit is raised | ~30 tok/s | lms get mlx-community/Qwen3.8-Flash-Next-4bit |
| M4 Max 128GB | LM Studio | MLX 4bit | 113.9 GB | fits if the Mac GPU limit is raised | ~41 tok/s | lms get mlx-community/Qwen3.8-Flash-Next-4bit |
| Mac Studio M4 Max (40c GPU) 128GB | LM Studio | MLX 4bit | 113.9 GB | fits if the Mac GPU limit is raised | ~41 tok/s | lms get mlx-community/Qwen3.8-Flash-Next-4bit |
| Mac Studio M5 Max (40c GPU) 128GB | LM Studio | MLX 4bit | 113.9 GB | fits if the Mac GPU limit is raised | ~46 tok/s | lms get mlx-community/Qwen3.8-Flash-Next-4bit |
| M2 Ultra 192GB | Ollama | Q4_K_M | 122.3 GB | fits in fast memory | ~56 tok/s | ollama run qwen3.8-flash-next:125b-a6b-q4_K_M |
| 2x Mac Studio M5 Max (40c GPU) 128GB | MLX | MLX 4bit | 113.9 GB | fits in fast memory | ~64 tok/s | exo # on every Mac (brew install --cask exo); load mlx-community/Qwen3.8-Flash-Next-4bit with Tensor sharding + RDMA |
| Mac Studio M3 Ultra 256GB | Ollama | Q8_0 | 191.3 GB | fits in fast memory | ~36 tok/s | ollama run qwen3.8-flash-next:125b-a6b-q8_0 |
| Mac Studio M5 Ultra (64c GPU) 256GB | Ollama | Q8_0 | 191.3 GB | fits in fast memory | ~53 tok/s | ollama run qwen3.8-flash-next:125b-a6b-q8_0 |
| Mac Studio M5 Ultra (80c GPU) 256GB | Ollama | Q8_0 | 191.3 GB | fits in fast memory | ~53 tok/s | ollama run qwen3.8-flash-next:125b-a6b-q8_0 |
| 2x DGX Spark 128GB | llama.cpp | Q8_0 GGUF | 201.6 GB | fits in fast memory | ~14 tok/s | llama-server -hf unsloth/Qwen3.8-Flash-Next-GGUF:Q8_0 --rpc node2:50052 -c 8192 # run rpc-server on each other node |
| 4x Mac Studio M5 Max (40c GPU) 128GB | llama.cpp | Q8_0 GGUF | 201.6 GB | fits in fast memory | ~19 tok/s | llama-server -hf unsloth/Qwen3.8-Flash-Next-GGUF:Q8_0 --rpc node2:50052,node3:50052,node4:50052 -c 8192 # run rpc-server on each other node |
| 2x Mac Studio M5 Ultra (80c GPU) 256GB | llama.cpp | Q8_0 GGUF | 201.6 GB | fits in fast memory | ~43 tok/s | llama-server -hf unsloth/Qwen3.8-Flash-Next-GGUF:Q8_0 --rpc node2:50052 -c 8192 # run rpc-server on each other node |
| Mac Studio M3 Ultra (80c GPU) 512GB | Ollama | Q8_0 | 191.3 GB | fits in fast memory | ~36 tok/s | ollama run qwen3.8-flash-next:125b-a6b-q8_0 |
| 4x DGX Spark 128GB | llama.cpp | Q8_0 GGUF | 201.6 GB | fits in fast memory | ~12 tok/s | llama-server -hf unsloth/Qwen3.8-Flash-Next-GGUF:Q8_0 --rpc node2:50052,node3:50052,node4:50052 -c 8192 # run rpc-server on each other node |
| 4x Mac Studio M5 Ultra (80c GPU) 256GB | llama.cpp | Q8_0 GGUF | 201.6 GB | fits in fast memory | ~38 tok/s | llama-server -hf unsloth/Qwen3.8-Flash-Next-GGUF:Q8_0 --rpc node2:50052,node3:50052,node4:50052 -c 8192 # run rpc-server on each other node |
Estimates, not benchmarks. Model sizes are ollama.com download sizes; memory needed adds headroom for ~8k tokens of context; speeds are rough figures from memory bandwidth. Data updated 2026-10-16.
A project by OnDevice Labs · llms.txt