will it local
256 GB/s quad LPDDR5X-8000; 32-128GB configs, up to 96GB GPU-accessible 8B Q4 ~22-30 tok/s; 32B-70B Q4 possible on 64GB+ configs (16-24 tok/s reported for 32B-class) — the non-Apple unified-memory path Vulkan/ROCm llama.cpp both work; ROCm for iGPU still fiddly — Vulkan is the pragmatic backend
Shared-memory device — speed scales with RAM bandwidth (256.0 GB/s shared RAM), usable for inference ≈ 30.0 GB of 32 GB system RAM
→explore the full catalogupgrade paths with live prices · what fits on each machinecomputed band 30.0-42.0 tok/s for 8B Q4_K_M (roofline, 256.0 GB/s shared RAM); community: 42.0 tok/s Shisa V2 8B Q4_K_M (llama.cpp Vulkan); 48.97-55.73 tok/s Llama 2 7B Q4_0 (ROCm FA / Vulkan)
Effective decode window: 0.60–0.85 of 256 GB/s nominal → ~154–218 GB/s effective (llama.cpp decode, Q4_K_M basis; arch window, community-calibrated; nominal is LPDDR/JEDEC channel arithmetic (shared memory))
→ 13b-q4 full 14.6 GB needed of 30 GB usable — headroom for context
→ 3b-q4 full 2.8 GB needed of 30 GB usable — headroom for context
→ 4b-q4 full 3.7 GB needed of 30 GB usable — headroom for context
→ 8b-q4 full 6.1 GB needed of 30 GB usable — headroom for context
→ 42.0 tok/s · Shisa V2 8B Q4_K_M llama.cpp Vulkan (RADV, 128GB LPDDR5X-8000) · forum.level1techs.com/t/233796 (lhl)
→ 48.97 tok/s · Llama 2 7B Q4_0 llama.cpp ROCm (FA) · llama.cpp disc#15021
→ 55.73 tok/s · Llama 2 7B Q4_0 llama.cpp Vulkan (RADV) · kyuz0.github.io/amd-strix-halo-toolboxes
Nominal bandwidth: 256.0 GB/s shared memory — see the effective decode window above
→ qwen3:0.6b on Tesla M60 (16GB, used) ($35) 0.6B · 0.6GB · needs ~1GB (4k ctx)
→ qwen3:1.7b on Tesla M60 (16GB, used) ($35) 1.7B · 1.4GB · needs ~1.8GB (4k ctx)
→ qwen3:4b on Tesla M60 (16GB, used) ($35) 4B · 2.6GB · needs ~3.2GB (4k ctx)
→ llama3.2:1b on Tesla M60 (16GB, used) ($35) 1B · 1.3GB · needs ~1.4GB (4k ctx)
→ llama3.2:3b on Tesla M60 (16GB, used) ($35) 3B · 2.0GB · needs ~2.4GB (4k ctx)
→ gemma3:1b on Tesla M60 (16GB, used) ($35) 1B · 1.0GB · needs ~1.1GB (4k ctx)
→ phi4:mini on Tesla M60 (16GB, used) ($35) mini 3.8B · 2.5GB · needs ~3GB (4k ctx)
→ moondream:2b on Tesla M60 (16GB, used) ($35) 2B · 1.7GB · needs ~3.7GB (4k ctx)
→ whisper:tiny on Tesla M60 (16GB, used) ($35) tiny · 0.1GB · needs ~2.1GB (4k ctx)
→ whisper on Tesla M60 (16GB, used) ($35) base · 0.3GB · needs ~2.3GB (4k ctx)
→ whisper:small on Tesla M60 (16GB, used) ($35) small · 0.9GB · needs ~2.9GB (4k ctx)
→ whisper:medium on Tesla M60 (16GB, used) ($35) medium · 3.1GB · needs ~5.1GB (4k ctx)
→ kokoro:82m on Tesla M60 (16GB, used) ($35) 82M · 0.3GB · needs ~2.3GB (4k ctx)
→ gemma3:4b on Tesla M60 (16GB, used) ($35) 4B · 3.3GB · needs ~3.8GB (4k ctx)
→ deepseek-r1:7b on Tesla M60 (16GB, used) ($35) distill 7B · 4.7GB · needs ~4.9GB (4k ctx)
→ mistral:7b on Tesla M60 (16GB, used) ($35) 7B · 4.1GB · needs ~4.6GB (4k ctx)
→ llava:7b on Tesla M60 (16GB, used) ($35) 7B · 4.7GB · needs ~6.7GB (4k ctx)
→ qwen3:8b on Tesla M60 (16GB, used) ($35) 8B · 5.2GB · needs ~5.8GB (4k ctx)
→ qwen2.5vl:7b on Tesla M60 (16GB, used) ($35) 7B · 5.6GB · needs ~5.8GB (4k ctx)
→ llama3.1:8b on Tesla M60 (16GB, used) ($35) 8B · 4.9GB · needs ~5.4GB (4k ctx)
→ deepseek-r1:8b on Tesla M60 (16GB, used) ($35) distill 8B · 4.9GB · needs ~5.4GB (4k ctx)
→ whisper:large-v3 on Tesla M60 (16GB, used) ($35) large-v3 · 6.2GB · needs ~8.2GB (4k ctx)
→ mistral-nemo:12b on Tesla M60 (16GB, used) ($35) 12B · 7.1GB · needs ~7.7GB (4k ctx)
→ sdxl on Tesla M60 (16GB, used) ($35) SDXL base · 6.9GB · needs ~8.9GB (4k ctx)
→ llava:13b on Tesla M60 (16GB, used) ($35) 13B · 8.0GB · needs ~11.1GB (4k ctx)
→ gemma3:12b on Tesla M60 (16GB, used) ($35) 12B · 8.1GB · needs ~9.6GB (4k ctx)
→ qwen3:14b on Tesla M60 (16GB, used) ($35) 14B · 9.3GB · needs ~9.9GB (4k ctx)
→ deepseek-r1:14b on Tesla M60 (16GB, used) ($35) distill 14B · 9.0GB · needs ~9.8GB (4k ctx)
→ phi4:14b on Tesla M60 (16GB, used) ($35) 14B · 9.1GB · needs ~9.9GB (4k ctx)
→ gpt-oss:20b on Tesla M40 (24GB, used) ($50) 20B MoE · 14GB · needs ~14.2GB (4k ctx)
→ devstral:24b on Tesla M40 (24GB, used) ($50) 24B · 14.6GB · needs ~15.2GB (4k ctx)
→ qwen3.8:27b on Tesla V100 32GB (used) ($700) 27B · 16.5GB · needs ~18.5GB (4k ctx)
→ gemma3:27b on Tesla V100 32GB (used) ($700) 27B · 17GB · needs ~18.9GB (4k ctx)
→ flux:schnell on Tesla V100 32GB (used) ($700) schnell fp8 · 17GB · needs ~19GB (4k ctx)
→ qwen3:30b-a3b on Tesla V100 32GB (used) ($700) 30B-A3B MoE · 18.6GB · needs ~19GB (4k ctx)
→ qwen3-coder:30b-a3b on Tesla V100 32GB (used) ($700) 30B-A3B MoE · 18.6GB · needs ~19GB (4k ctx)
→ qwen3:32b on Tesla V100 32GB (used) ($700) 32B · 20GB · needs ~21GB (4k ctx)
→ deepseek-r1:32b on Tesla V100 32GB (used) ($700) distill 32B · 20GB · needs ~21GB (4k ctx)
→ qwen2.5vl:32b on Mac mini M5 Pro (64GB) ($1669) 32B · 22GB · needs ~23GB (4k ctx)
→ flux:dev on Mac mini M5 Pro (64GB) ($1669) dev fp8 · 23GB · needs ~25GB (4k ctx)
→ mixtral:8x7b on Mac mini M5 Pro (64GB) ($1669) 8x7B MoE · 26GB · needs ~26.5GB (4k ctx)
—
⚠ 256 GB/s quad LPDDR5X-8000; 32-128GB configs, up to 96GB GPU-accessible
⚠ 8B Q4 ~22-30 tok/s; 32B-70B Q4 possible on 64GB+ configs (16-24 tok/s reported for 32B-class) — the non-Apple unified-memory path
⚠ Vulkan/ROCm llama.cpp both work; ROCm for iGPU still fiddly — Vulkan is the pragmatic backend
—