Skip to main content
Watch: Grok Bot Built My AI Video Arena While I Tested 8 Models

Memory tier - verified

Open models that fit in 64 GB

33 open-weight models of 63 in this directory fit in 64 GB of memory, sorted by total parameters, largest first.

Devices with 64 GB: MacBook Pro (M5 Pro), Mac mini (M5 Pro), MacBook Pro (M5 Max), Mac Studio (M5 Max) and DGX Spark. Check your GPU or Mac across 33 devices.

33 models

Total params, largest first
33 open-weight models that fit in 64 GB, by total parameters, largest first
CommercialOllama
Qwen3-Coder-NextQwen/Qwen3-Coder-NextAlibaba Cloud (Qwen)79.7B3B active44.8 GB4bit MLX262KCommercial use: Yesqwen3-coder-next
NVIDIA Nemotron 3.5 Bolt 74B-A9Bnvidia/NVIDIA-Nemotron-3.5-Bolt-74B-A9B-BF16NVIDIA74.9B9B active22.0 GB24GB (native 2-bit) GGUF262KCommercial use: Yesnemotron-3.5-bolt:74b-a9b
AI21 Jamba2 Miniai21labs/AI21-Jamba2-MiniAI21 Labs51.6B12B active31.2 GBQ4_K_M GGUF262KCommercial use: Yes-No Ollama tag
Qwen3.6-35B-A3BQwen/Qwen3.6-35B-A3BAlibaba Cloud (Qwen)36B3B active20.4 GB4bit MLX262KCommercial use: Yesqwen3.6:35b
Cohere Aya Expanse 32BCohereLabs/aya-expanse-32bCohere32.3B19.8 GBQ4_K_M GGUF131KCommercial use: Noaya-expanse:32b
Ai2 Olmo 3.1 32B Thinkallenai/Olmo-3.1-32B-ThinkAi232.2B18.1 GB4bit MLX66KCommercial use: Yesolmo-3.1:32b-think
NVIDIA Nemotron 3.5 Lightning 30B-A3Bnvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16NVIDIA31.6B3B active17.8 GB4bit MLX1.05MCommercial use: Yesnemotron-3.5-lightning:30b
Gemma 4 31Bgoogle/gemma-4-31B-itGoogle31.3B17.7 GBQ4_0 (QAT) GGUF262KCommercial use: Yesgemma4:31b
IBM Granite 4.2 30Bibm-granite/granite-4.2-30bIBM29.3B16.5 GB4bit MLX131KCommercial use: Yesgranite4.2:30b
Qwen3.8-27BQwen/Qwen3.8-27BAlibaba Cloud (Qwen)27.8B16.1 GB4bit MLX262KCommercial use: Yesqwen3.8:27b
Gemma 4 26B A4Bgoogle/gemma-4-26B-A4B-itGoogle25.8B3.8B active14.4 GBQ4_0 (QAT) GGUF262KCommercial use: Yesgemma4:26b
Devstral Small 2 24B Instruct 2512mistralai/Devstral-Small-2-24B-Instruct-2512Mistral AI24B14.3 GBQ4_K_M GGUF262KCommercial use: Yesdevstral-small-2:24b
Liquid AI LFM2.5-24B-A2BLiquidAI/LFM2.5-24B-A2BLiquid AI24B2.5B active13.6 GB4bit MLX131KCommercial use: Conditionslfm2.5:24b
gpt-oss-20bopenai/gpt-oss-20bOpenAI20.9B3.6B active11.2 GBMXFP4-Q4 MLX131KCommercial use: Yesgpt-oss:20b
Phi-4-reasoning-vision-15Bmicrosoft/Phi-4-reasoning-vision-15BMicrosoft15.1B30.2 GBOfficial BF1616KCommercial use: Yes-No Ollama tag
Phi-4microsoft/phi-4Microsoft14.7B8.2 GB4bit MLX16KCommercial use: Yesphi4:14b
Phi-4-reasoning-plusmicrosoft/Phi-4-reasoning-plusMicrosoft14.7B9.1 GBQ4_K_M GGUF33KCommercial use: Yesphi4-reasoning:plus
Ministral 3 14B Instruct 2512mistralai/Ministral-3-14B-Instruct-2512Mistral AI13.9B8.2 GBQ4_K_M GGUF262KCommercial use: Yesministral-3:14b
Gemma 4 12Bgoogle/gemma-4-12B-itGoogle12B6.7 GB4bit MLX262KCommercial use: Yesgemma4:12b
Qwen3.5-9BQwen/Qwen3.5-9BAlibaba Cloud (Qwen)9.7B5.7 GBQ4_K_M GGUF262KCommercial use: Yesqwen3.5:9b
Ministral 3 8B Instruct 2512mistralai/Ministral-3-8B-Instruct-2512Mistral AI8.9B5.2 GBQ4_K_M GGUF262KCommercial use: Yesministral-3:8b
IBM Granite 4.2 8Bibm-granite/granite-4.2-8bIBM8.8B4.9 GB4bit MLX131KCommercial use: Yesgranite4.2:8b
Cohere Command R7B (12-2024)CohereLabs/c4ai-command-r7b-12-2024Cohere8B4.5 GB4bit MLX131KCommercial use: Nocommand-r7b:7b
Gemma 4 E4Bgoogle/gemma-4-E4B-itGoogle8B5.0 GBQ4_K_M GGUF131KCommercial use: Yesgemma4:e4b
Ai2 Olmo Hybrid Instruct DPO 7Ballenai/Olmo-Hybrid-Instruct-DPO-7BAi27.4B14.9 GBOfficial BF1633KCommercial use: Yes-No Ollama tag
Gemma 4 E2Bgoogle/gemma-4-E2B-itGoogle5.1B3.1 GBQ4_K_M GGUF131KCommercial use: Yesgemma4:e2b
Qwen3.5-4BQwen/Qwen3.5-4BAlibaba Cloud (Qwen)4.7B2.7 GBQ4_K_M GGUF262KCommercial use: Yesqwen3.5:4b
Ministral 3 3B Instruct 2512mistralai/Ministral-3-3B-Instruct-2512Mistral AI3.8B2.1 GBQ4_K_M GGUF262KCommercial use: Yesministral-3:3b
Phi-4-mini-instructmicrosoft/Phi-4-mini-instructMicrosoft3.8B2.2 GB4bit MLX131KCommercial use: Yesphi4-mini:3.8b
Phi-4-mini-reasoningmicrosoft/Phi-4-mini-reasoningMicrosoft3.8B2.2 GB4bit MLX131KCommercial use: Yesphi4-mini-reasoning:3.8b
Liquid AI LFM2.5-2.6BLiquidAI/LFM2.5-2.6BLiquid AI2.7B1.6 GB4bit MLX131KCommercial use: Conditions-No Ollama tag
Qwen3.5-2BQwen/Qwen3.5-2BAlibaba Cloud (Qwen)2.3B1.3 GBQ4_K_M GGUF262KCommercial use: Yesqwen3.5:2b
Qwen3.5-0.8BQwen/Qwen3.5-0.8BAlibaba Cloud (Qwen)0.9B0.5 GBQ4_K_M GGUF262KCommercial use: Yesqwen3.5:0.8b

More memory? See what fits in 128 GB or browse every local model.

FAQ

How many open models fit in 64 GB?

33 of 63 models in this directory have a listed download of 54.4 GB or less, 85% of 64 GB.

What is the largest model that fits in 64 GB?

By total parameters, Qwen3-Coder-Next (79.7B), whose smallest listed download is 44.8 GB, the 4bit MLX build from mlx-community/Qwen3-Coder-Next-4bit. Largest is a size, not a quality ranking.

Which of these can I use commercially?

29 of the 33 ship under licenses that allow commercial use outright. The license column shows the rest: some allow it with conditions and some do not.

Why 85% of 64 GB?

The download is only the weights. Running a model also needs memory for the context (the KV cache grows with every token you keep) and for the operating system and other apps, so this site leaves 15% free as a rule of thumb. A short context needs less; a long one can need more.

Newsletter

New open weights, every week

One email per week. New model releases, open-weight drops and what they mean for what you build.