Memory budget · 48 GB
Best local LLMs for 48GB
The Apple M5 Pro (48GB) gives about ~32 GB of its 48GB to model weights after the OS takes its share. 129 of 161 local LLMs fit, biggest first.
- Usable range
- 32 GB
- Models that fit
- 129
- Memory types
- 1
- Top pick
- 39.8B
What 48GB actually gives you
Usable figures are sourced per device (tap a card for the full profile). Verdicts below use Q4_K_M, the community-default quant.
Every tracked 48GB device
Most usable memory first Sourced specs
Each device has its own profile with the full ranked list. Memory bandwidth, where listed, sets how fast tokens generate; it does not change what fits.
Top pick for 48GB Q4_K_M
Flint 40B-A2.5B 39.8B MoE
Runs comfortably on the most capable 48GB setup (Apple M5 Pro (48GB), ~32 GB usable) at ~25.9 GB. Check it against your exact device on its model page.
Models ranked for 48GB
Biggest that fits first Mac
- NELlama-3.3-Nemotron-Super-49B-v149B · ~30.6 GB at Q4_K_M
- Mixtral 8x7B46.7B MoE · ~28.9 GB at Q4_K_M
- FLFlint 40B-A2.5B39.8B MoE · ~25.9 GB at Q4_K_M
- SESeed-OSS 36B Instruct36B · ~22.5 GB at Q4_K_M
- Qwen3.6 35B-A3B36B MoE · ~24.5 GB at Q4_K_M
- CRCommand R 35B35B · ~22.3 GB at Q4_K_M
- OROrnith 1.0 35B35B MoE · ~23.2 GB at Q4_K_M
- OROrnith 1.5 35B-A3B35B MoE · ~22.4 GB at Q4_K_M
- Qwen-AgentWorld 35B-A3B34.7B MoE · ~24.3 GB at Q4_K_M
- YiYi 1.5 34B34B · ~21.4 GB at Q4_K_M
- FNFalcon-H1-34B-Instruct34B · ~21.1 GB at Q4_K_M
- LALaguna XS 2.133.4B MoE · ~22.2 GB at Q4_K_M
- Gemma 4 31B32.7B · ~20.5 GB at Q4_K_M
- Qwen2.5 32B32B · ~22.1 GB at Q4_K_M
- DeepSeek-R1-Distill-Qwen 32B32B · ~22.1 GB at Q4_K_M
- Qwen2.5 Coder 32B32B · ~20.7 GB at Q4_K_M
- Granite 4.0 H Small32B MoE · ~20.4 GB at Q4_K_M
- GLGLM-4-32B-041432B · ~20.5 GB at Q4_K_M
- EXEXAONE 4.0 32B32B · ~20.2 GB at Q4_K_M
- OLOLMo 2 32B Instruct32B · ~20.3 GB at Q4_K_M
- OLOlmo 3.1 32B Instruct32B · ~20.3 GB at Q4_K_M
- Qwen3-VL 30B-A3B31.1B MoE · ~20.4 GB at Q4_K_M
- Qwen3 30B-A3B30.5B MoE · ~20.7 GB at Q4_K_M · Elo 1383
- Qwen3-Coder 30B-A3B30.5B MoE · ~19.4 GB at Q4_K_M
- NONorth Mini Code 1.030.5B MoE · ~21.1 GB at Q4_K_M
- SSarvam-30B30B MoE · ~21.7 GB at Q4_K_M
- NENemotron 3 Nano 30B-A3B30B MoE · ~25.1 GB at Q4_K_M
- NENemotron Cascade 2 30B-A3B30B MoE · ~25.1 GB at Q4_K_M
- GLGLM-4.7-Flash30B MoE · ~19.2 GB at Q4_K_M
- MUMuse Glimmer 30B30B · ~18.2 GB at Q4_K_M
- NENemotron 3.5 Lightning 30B-A3B30B MoE · ~25.8 GB at Q4_K_M
- Granite 4.2 30B30B · ~18.6 GB at Q4_K_M
- Granite 4.1 30B28.9B · ~19.1 GB at Q4_K_M
- Qwen3.6 27B27.8B · ~18.9 GB at Q4_K_M
- Gemma 2 27B27B · ~18.7 GB at Q4_K_M · Elo 1289
- Gemma 3 27B27B · ~18.6 GB at Q4_K_M · Elo 1366
- Qwen3.8 27B27B · ~17.3 GB at Q4_K_M
- Gemma 4 26B-A4B26.5B MoE · ~19 GB at Q4_K_M
- Mistral Small 3 24B24B · ~16.3 GB at Q4_K_M · Elo 1357
- SSarvam-M 24B24B · ~16.3 GB at Q4_K_M
- Mistral Small 3.1 24B24B · ~15.4 GB at Q4_K_M
- Magistral Small24B · ~16 GB at Q4_K_M
- Devstral Small24B · ~15.4 GB at Q4_K_M
- LFLFM2 24B-A2B24B MoE · ~15.4 GB at Q4_K_M
- Devstral Small 2 24B24B · ~15.4 GB at Q4_K_M
- gpt-oss 20B21B MoE · ~13.2 GB at Q4_K_M
- ERERNIE 4.5 21B-A3B21B MoE · ~14.3 GB at Q4_K_M
- DeepSeek-V2-Lite16B MoE · ~12.2 GB at Q4_K_M
- Qwen2.5 14B14B · ~10.7 GB at Q4_K_M
- DeepSeek-R1-Distill-Qwen 14B14B · ~10.7 GB at Q4_K_M
- Qwen2.5 Coder 14B14B · ~10.1 GB at Q4_K_M
- Phi-4-reasoning14B · ~10.1 GB at Q4_K_M
- Ministral 3 14B14B · ~9.4 GB at Q4_K_M
- Mistral Nemo 12B12.2B · ~8.6 GB at Q4_K_M
- Gemma 3 12B12B · ~8.9 GB at Q4_K_M · Elo 1342
- Gemma 4 12B12B · ~8.7 GB at Q4_K_M
- Llama 3.2 Vision 11B10.7B · ~9 GB at Q4_K_M
- FNFalcon3 10B10B · ~7.5 GB at Q4_K_M
- Gemma 2 9B9B · ~7.3 GB at Q4_K_M · Elo 1266
- GLGLM-4-9B-04149B · ~7.2 GB at Q4_K_M
- NENemotron Nano 9B v29B · ~7.6 GB at Q4_K_M
- OROrnith 1.0 9B9B · ~7.1 GB at Q4_K_M
- OROrnith 1.5 9B9B · ~6.9 GB at Q4_K_M
- Granite 4.1 8B8.8B · ~6.8 GB at Q4_K_M
- Qwen3-VL 8B8.77B · ~7.3 GB at Q4_K_M
- LFLFM2.5 8B-A1B8.3B MoE · ~6.3 GB at Q4_K_M
- Qwen2.5-VL 7B8.29B · ~7.1 GB at Q4_K_M
- DeepSeek-R1-0528-Qwen3-8B8.19B · ~6.2 GB at Q4_K_M
- Llama 3.1 8B8B · ~6.4 GB at Q4_K_M · Elo 1211
- DeepSeek-R1-Distill-Llama 8B8B · ~6.4 GB at Q4_K_M
- Gemma 3n E4B8B · ~5.7 GB at Q4_K_M
- Gemma 4 E4B8B · ~6.5 GB at Q4_K_M
- Ministral 3 8B8B · ~6.3 GB at Q4_K_M
- Granite 4.2 8B8B · ~6.5 GB at Q4_K_M
- NENemotron 3.5 Bolt E4B8B · ~8.9 GB at Q4_K_M
- Mistral 7B7B · ~5.8 GB at Q4_K_M · Elo 1149
- Qwen2.5 7B7B · ~6.1 GB at Q4_K_M
- DeepSeek-R1-Distill-Qwen 7B7B · ~6.1 GB at Q4_K_M
- Qwen2.5 Coder 7B7B · ~5.8 GB at Q4_K_M
- OLOlmo 3 7B Instruct7B · ~5.6 GB at Q4_K_M
- Gemma 4 E2B5.1B · ~4.4 GB at Q4_K_M
- Qwen3-VL 4B4.44B · ~4.4 GB at Q4_K_M
- Gemma 3 4B4B · ~3.8 GB at Q4_K_M · Elo 1303
- NENemotron 3 Nano 4B4B · ~3.9 GB at Q4_K_M
- Phi-3.5-mini 3.8B3.82B · ~3.7 GB at Q4_K_M
- Phi-4-mini 3.8B3.8B · ~3.8 GB at Q4_K_M
- Phi-4-mini-reasoning3.8B · ~3.6 GB at Q4_K_M
- Qwen2.5-VL 3B3.75B · ~4.4 GB at Q4_K_M
- Granite 4.1 3B3.4B · ~3.3 GB at Q4_K_M
- Qwen2.5 3B3.09B · ~3.3 GB at Q4_K_M
- Qwen2.5 Coder 3B3.09B · ~3 GB at Q4_K_M
- Llama 3.2 3B3B · ~3.2 GB at Q4_K_M · Elo 1166
- SmolLM3 3B3B · ~3 GB at Q4_K_M
- OPApple OpenELM 3B3B · ~3 GB at Q4_K_M
- Ministral 3 3B3B · ~3.2 GB at Q4_K_M
- Granite 4.2 3B3B · ~3.3 GB at Q4_K_M
- Gemma 2 2B2.61B · ~2.9 GB at Q4_K_M
- Granite 3.1 2B2.53B · ~2.8 GB at Q4_K_M
- LFLFM3 Omni 2B2.17B · ~3.1 GB at Q4_K_M
- SSarvam-1 2B2B · ~2.7 GB at Q4_K_M
- LFLFM3-VL-2B1.92B · ~2.5 GB at Q4_K_M
- SmolLM2 1.7B1.7B · ~2.2 GB at Q4_K_M · Elo 1114
- Qwen3 1.7B1.7B · ~2.4 GB at Q4_K_M
- Qwen2.5 1.5B1.54B · ~2.2 GB at Q4_K_M
- Qwen2.5 Coder 1.5B1.54B · ~2 GB at Q4_K_M
- MIMiniCPM-V 4.61.3B · ~1.6 GB at Q4_K_M
- LFLFM2.5 1.2B1.17B · ~1.8 GB at Q4_K_M
- LFLFM2.5 1.2B Thinking1.17B · ~1.8 GB at Q4_K_M
- TLTinyLlama 1.1B1.1B · ~1.8 GB at Q4_K_M
- OPApple OpenELM 1.1B1.1B · ~1.7 GB at Q4_K_M
- Llama 3.2 1B1B · ~1.8 GB at Q4_K_M · Elo 1110
- Gemma 3 1B1B · ~1.8 GB at Q4_K_M
- Qwen3 0.6B0.6B · ~1.5 GB at Q4_K_M
- Qwen2.5 0.5B0.494B · ~1.5 GB at Q4_K_M
- Qwen2.5 Coder 0.5B0.494B · ~1.4 GB at Q4_K_M
- SmolLM2 360M0.362B · ~1.2 GB at Q4_K_M
- Gemma 3 270M0.27B · ~1.1 GB at Q4_K_M
- SmolLM2 135M0.135B · ~1 GB at Q4_K_M
Each chip links to the full breakdown for that model on a real 48GB device. "Tight" means it fits but with little headroom, close other apps.
The ceiling, per memory type
Apple M5 Pro (48GB) (~32 GB usable)
Runs up to Flint 40B-A2.5B (39.8B) comfortably at Q4_K_M. Larger models either sit tight or spill past the ~32 GB it can give a model.
Too large for any 48GB device
Llama 3.3 70B 70B Qwen2.5 72B 72B Hunyuan-A13B-Instruct 80B Qwen3-Coder-Next 80B-A3B 80B Sarvam-105B 105B GLM-4.5-Air 106B Llama 4 Scout 109B Command A 111B gpt-oss 120B 117B Laguna S 2.1 118B Nemotron 3 Super 120B-A12B 120B Nemotron 3.5 Super VL 120B-A12B 120B Devstral 2 123B 123B dots.llm1 142B Qwen3 235B A22B 235B DeepSeek-V4-Flash 284B GLM-5.3-Flash 320B GLM-4.6 357B Llama 4 Maverick 400B MiniMax M3 428B MiniMax-M1-80k 456B Qwen3-Coder 480B-A35B Instruct 480B Nemotron 3 Ultra 550B-A55B 550B DeepSeek R1 671B DeepSeek V3 671B DeepSeek-R1-0528 671B GLM-5.2 744B Kimi K2 Instruct 1000B Kimi K2.6 1000B Kimi K2.7 Code 1000B DeepSeek-V4-Pro 1600B Kimi K3 2800B
FAQ
How much of 48GB can a model actually use?
It depends on the memory type. Apple unified memory: about 32 GB (Apple M5 Pro (48GB)). The rest is reserved for the OS, display and runtime overhead.
What is the best local LLM for 48GB?
Flint 40B-A2.5B (39.8B) is the strongest model that runs comfortably at Q4_K_M on the most capable 48GB setup (Apple M5 Pro (48GB), ~32 GB usable). On a tighter 48GB device the ceiling is lower, shown per row above.
Sources
-
apple.com · 2 sources
-
blog.peddals.com · 1 source
-
huggingface.co · 3 sources
-
ollama.com · 1 source
-
support.apple.com · 1 source
Memory figures are estimates at Q4_K_M with a small context. See methodology.