Best Local LLM by Mac RAM

According to the LLMCheck index, the best local LLM for a 16 GB Mac is Gemma 4 12B at ~17 tok/s (estimated) on a Mac mini M6; for 32 GB, Qwen 3.6-35B-A3B at 64 tok/s (community run) on a Mac mini M6; for 128 GB, Qwen3.8-Flash-Next at 44 tok/s (community run) on a Mac Studio M5 Max.

Jump to a tier: 8 GB · 16 GB · 24 GB · 32 GB · 64 GB · 128 GB · 192–256 GB · 512 GB
Know your exact Mac? Pick it from every config → · Buying? The Mac Advisor names one config →

Rule of thumb: a 4-bit model needs about 0.57 GB per billion parameters plus ~2 GB, and macOS needs its own share. Mixture-of-experts models read only their active parameters per token, so they run far faster than a dense model of the same size. Speeds come from the index’s bandwidth model unless marked as a published figure or community run.

Best local LLM for a 8 GB Mac

21 of the 111 models in the LLMCheck index fit in 8 GB. The top pick is Nanbeige4.2-3B at ~30 tok/s (estimated), the best-ranked model that reaches 15 tok/s here; speeds are for a MacBook Air M1 (8 GB) (68 GB/s), and a faster chip with the same memory runs every model faster.

⭐ Best pick

Nanbeige4.2-3B

3B · Apache 2.0 · cap 20/50

estimated30 tok/s

⚡ Fastest

Gemma 4 E2B

2B · Apache 2.0 · cap 13/50

estimated44 tok/s

🧠 Most capable

Bonsai 27B

27B · Apache 2.0 · cap 33/50

—

#	Model	Size	License	Speed	Capability
1	Nanbeige4.2-3B	3B	Apache 2.0	estimated30 tok/s	20/50
2	Bonsai 27B	27B	Apache 2.0	—	33/50
3	MiMo-V2.6-Distill-Qwen-9B slow	9.41B	MIT	estimated10 tok/s	21/50
4	Maple Preview 20B-A1B	20B MoE	MIT	—	20/50
5	Qwen 3.5 9B slow	9B	Apache 2.0	estimated10 tok/s	18/50
6	Gemma 4 E4B	4B	Apache 2.0	estimated23 tok/s	16/50
7	Gemma 4 E2B	2B	Apache 2.0	estimated44 tok/s	13/50
8	LFM2.5-2.6B	2.6B	LFM Open	estimated34 tok/s	14/50

Every Mac in this tier (8 GB): MacBook Air M1 (8 GB)

Best local LLM for a 16 GB Mac

32 of the 111 models in the LLMCheck index fit in 16 GB. The top pick is Gemma 4 12B at ~17 tok/s (estimated), the best-ranked model that reaches 15 tok/s here; speeds are for a Mac mini M6 (16 GB) (153 GB/s), and a faster chip with the same memory runs every model faster.

⭐ Best pick

Gemma 4 12B

12B · Apache 2.0 · cap 30/50

estimated17 tok/s

⚡ Fastest

Gemma 4 E2B

2B · Apache 2.0 · cap 13/50

estimated91 tok/s

🧠 Most capable

Ternary Bonsai 2 27B

27.36B · Apache 2.0 · cap 37/50

—

#	Model	Size	License	Speed	Capability
1	Gemma 4 12B	12B	Apache 2.0	estimated17 tok/s	30/50
2	Ternary Bonsai 2 27B	27.36B	Apache 2.0	—	37/50
3	Bonsai 27B	27B	Apache 2.0	—	33/50
4	Nanbeige4.2-3B	3B	Apache 2.0	estimated64 tok/s	20/50
5	GPT-oss 20B	21B MoE	Apache 2.0	estimated41 tok/s	22/50
6	Mistral Small 3.2 24B slow	24B	Apache 2.0	estimated8 tok/s	25/50
7	Devstral Small 24B slow	24B	Apache 2.0	estimated8 tok/s	24/50
8	Mellum2.1-12B-A2.5B-Thinking	12.15B MoE	Apache 2.0	estimated55 tok/s	18/50

Every Mac in this tier (16 GB): Mac mini M6 (16 GB) · Mac mini M4 (16 GB) · MacBook Air M4 (16 GB) · Mac mini M2 Pro (16 GB) · MacBook Air M2 (16 GB) · MacBook Air M3 (16 GB) · MacBook Air M1 (16 GB)

Best local LLM for a 24 GB Mac

54 of the 111 models in the LLMCheck index fit in 24 GB. The top pick is Qwen 3.6-35B-A3B at 64 tok/s (community run), the best-ranked model that reaches 15 tok/s here; speeds are for a Mac mini M6 (24 GB) (170 GB/s), and a faster chip with the same memory runs every model faster.

⭐ Best pick

Qwen 3.6-35B-A3B

35B MoE · Apache 2.0 · cap 38/50

community64 tok/s

⚡ Fastest

Gemma 4 E2B

2B · Apache 2.0 · cap 13/50

estimated99 tok/s

🧠 Most capable (slow)

Qwen3.8-27B

27.8B · Apache 2.0 · cap 46/50

community9 tok/s

#	Model	Size	License	Speed	Capability
1	Qwen 3.6-35B-A3B	35B MoE	Apache 2.0	community64 tok/s	38/50
2	Qwen3.8-27B slow	27.8B	Apache 2.0	community9 tok/s	46/50
3	Qwen 3.6-27B slow	27B	Apache 2.0	estimated9 tok/s	44/50
4	KAT-Coder-V2.5	35B MoE	Apache 2.0	estimated47 tok/s	39/50
5	LFM2.5-24B-A2B	24B MoE	LFM Open	estimated79 tok/s	35/50
6	Muse Glimmer 30B slow	30B	Apache 2.0	community9 tok/s	42/50
7	Nemotron 3.5 Lightning	30B MoE	OpenMDW	community45 tok/s	36/50
8	Gemma 4 31B slow	31B	Apache 2.0	community8 tok/s	40/50

Every Mac in this tier (24 GB): Mac mini M5 Pro (24 GB) · MacBook Pro M5 Pro (24 GB) · Mac mini M4 Pro (24 GB) · Mac mini M6 (24 GB) · Mac mini M4 (24 GB) · MacBook Air M4 (24 GB) · MacBook Air M2 (24 GB) · MacBook Air M3 (24 GB)

Best local LLM for a 32 GB Mac

54 of the 111 models in the LLMCheck index fit in 32 GB. The top pick is Qwen 3.6-35B-A3B at 64 tok/s (community run), the best-ranked model that reaches 15 tok/s here; speeds are for a Mac mini M6 (32 GB) (170 GB/s), and a faster chip with the same memory runs every model faster.

⭐ Best pick

Qwen 3.6-35B-A3B

35B MoE · Apache 2.0 · cap 38/50

community64 tok/s

⚡ Fastest

Gemma 4 E2B

2B · Apache 2.0 · cap 13/50

estimated99 tok/s

🧠 Most capable (slow)

Qwen3.8-27B

27.8B · Apache 2.0 · cap 46/50

community9 tok/s

#	Model	Size	License	Speed	Capability
1	Qwen 3.6-35B-A3B	35B MoE	Apache 2.0	community64 tok/s	38/50
2	Qwen3.8-27B slow	27.8B	Apache 2.0	community9 tok/s	46/50
3	Qwen 3.6-27B slow	27B	Apache 2.0	estimated9 tok/s	44/50
4	KAT-Coder-V2.5	35B MoE	Apache 2.0	estimated47 tok/s	39/50
5	LFM2.5-24B-A2B	24B MoE	LFM Open	estimated79 tok/s	35/50
6	Muse Glimmer 30B slow	30B	Apache 2.0	community9 tok/s	42/50
7	Nemotron 3.5 Lightning	30B MoE	OpenMDW	community45 tok/s	36/50
8	Gemma 4 31B slow	31B	Apache 2.0	community8 tok/s	40/50

Every Mac in this tier (32–48 GB): Mac mini M6 (32 GB) · Mac Studio M2 Max (32 GB) · Mac mini M2 Pro (32 GB) · Mac Studio M5 Max (36 GB) · Mac mini M5 Pro (48 GB)

Best local LLM for a 64 GB Mac

66 of the 111 models in the LLMCheck index fit in 64 GB. The top pick is Qwen 3.6-35B-A3B at ~92 tok/s (estimated), the best-ranked model that reaches 15 tok/s here; speeds are for a Mac mini M5 Pro (64 GB) (307 GB/s), and a faster chip with the same memory runs every model faster.

⭐ Best pick

Qwen 3.6-35B-A3B

35B MoE · Apache 2.0 · cap 38/50

estimated92 tok/s

⚡ Fastest

Maple Preview 20B-A1B

20B MoE · MIT · cap 20/50

sourced387 tok/s

🧠 Most capable

Qwen3.8-27B

27.8B · Apache 2.0 · cap 46/50

estimated15 tok/s

#	Model	Size	License	Speed	Capability
1	Qwen 3.6-35B-A3B	35B MoE	Apache 2.0	estimated92 tok/s	38/50
2	LFM2.5-24B-A2B	24B MoE	LFM Open	estimated109 tok/s	35/50
3	Qwen3.8-27B	27.8B	Apache 2.0	estimated15 tok/s	46/50
4	KAT-Coder-V2.5	35B MoE	Apache 2.0	estimated72 tok/s	39/50
5	Qwen 3.6-27B	27B	Apache 2.0	estimated15 tok/s	44/50
6	Nemotron 3.5 Bolt 74B-A9B	74B MoE	OpenMDW	estimated38 tok/s	40/50
7	North Mini Code 1.0	30.5B MoE	Apache 2.0	community70 tok/s	36/50
8	Gemma 4 26B-A4B	26B MoE	Apache 2.0	estimated78 tok/s	35/50

Every Mac in this tier (64–96 GB): Mac Studio M5 Max (64 GB) · MacBook Pro M5 Max (64 GB) · Mac Studio M4 Max (64 GB) · Mac mini M5 Pro (64 GB) · MacBook Pro M5 Pro (64 GB) · Mac mini M4 Pro (64 GB) · Mac Studio M1 Ultra (64 GB) · Mac Studio M2 Max (64 GB) · MacBook Pro M3 Max (64 GB) · Mac Studio M5 Ultra (96 GB)

Best local LLM for a 128 GB Mac

80 of the 111 models in the LLMCheck index fit in 128 GB. The top pick is Qwen3.8-Flash-Next at 44 tok/s (community run), the best-ranked model that reaches 15 tok/s here; speeds are for a Mac Studio M5 Max (128 GB) (614 GB/s), and a faster chip with the same memory runs every model faster.

⭐ Best pick

Qwen3.8-Flash-Next

125B MoE · Qwen Custom · cap 47/50

community44 tok/s

⚡ Fastest

Gemma 4 E2B

2B · Apache 2.0 · cap 13/50

estimated261 tok/s

🧠 Most capable

GLM-5.3-Flash

320B MoE · MIT · cap 48/50

estimated27 tok/s

#	Model	Size	License	Speed	Capability
1	Qwen3.8-Flash-Next	125B MoE	Qwen Custom	community44 tok/s	47/50
2	Qwen 3.6-35B-A3B	35B MoE	Apache 2.0	sourced118 tok/s	38/50
3	DeepSeek V4 Flash	284B MoE	MIT	community39 tok/s	47/50
4	DeepSeek V4 Flash Vision-Exp	305B MoE	MIT	estimated39 tok/s	47/50
5	GLM-5.3-Flash	320B MoE	MIT	estimated27 tok/s	48/50
6	Inkling-Small	276B MoE	Apache 2.0	estimated42 tok/s	46/50
7	North Mini Code 1.0	30.5B MoE	Apache 2.0	community120 tok/s	36/50
8	LFM2.5-24B-A2B	24B MoE	LFM Open	sourced184 tok/s	35/50

Every Mac in this tier (128 GB): Mac Studio M5 Max (128 GB) · MacBook Pro M5 Max (128 GB) · Mac Studio M4 Max (128 GB) · Mac Studio M1 Ultra (128 GB) · Mac Studio M2 Ultra (128 GB) · MacBook Pro M3 Max (128 GB)

Best local LLM for a 192–256 GB Mac

86 of the 111 models in the LLMCheck index fit in 192 GB, the smallest Mac in this tier. The top pick is MiMo-V2.6-Flash at ~79 tok/s (estimated), the best-ranked model that reaches 15 tok/s here; speeds are for a Mac Studio M5 Ultra (256 GB) (1228 GB/s), and a faster chip with the same memory runs every model faster.

⭐ Best pick

MiMo-V2.6-Flash

309B MoE · MIT · cap 49/50

estimated79 tok/s

⚡ Fastest

Gemma 4 E2B

2B · Apache 2.0 · cap 13/50

estimated378 tok/s

🧠 Most capable

MiniMax-M3

428B MoE · MiniMax Community · cap 48/50

estimated42 tok/s

#	Model	Size	License	Speed	Capability
1	MiMo-V2.6-Flash	309B MoE	MIT	estimated79 tok/s	49/50
2	Qwen3.8-Flash-Next	125B MoE	Qwen Custom	community57 tok/s	47/50
3	KAT-Coder-V2.5	35B MoE	Apache 2.0	estimated120 tok/s	39/50
4	MiniMax-M3	428B MoE	MiniMax Community	estimated42 tok/s	48/50
5	Qwen 3.6-35B-A3B	35B MoE	Apache 2.0	community153 tok/s	38/50
6	DeepSeek V4 Flash Vision-Exp	305B MoE	MIT	estimated46 tok/s	47/50
7	DeepSeek V4 Flash	284B MoE	MIT	estimated45 tok/s	47/50
8	GLM-5.3-Flash	320B MoE	MIT	estimated35 tok/s	48/50

Every Mac in this tier (192–256 GB): Mac Studio M2 Ultra (192 GB) · Mac Studio M5 Ultra (256 GB) · Mac Studio M3 Ultra (256 GB)

Best local LLM for a 512 GB Mac

104 of the 111 models in the LLMCheck index fit in 512 GB. The top pick is MiMo-V2.6-Flash at ~79 tok/s (estimated), the best-ranked model that reaches 15 tok/s here; speeds are for a Mac Studio M5 Ultra (512 GB) (1228 GB/s), and a faster chip with the same memory runs every model faster.

⭐ Best pick

MiMo-V2.6-Flash

309B MoE · MIT · cap 49/50

estimated79 tok/s

⚡ Fastest

Gemma 4 E2B

2B · Apache 2.0 · cap 13/50

estimated378 tok/s

🧠 Most capable

DeepSeek V4.1 Flash

552B MoE · MIT · cap 50/50

estimated29 tok/s

#	Model	Size	License	Speed	Capability
1	MiMo-V2.6-Flash	309B MoE	MIT	estimated79 tok/s	49/50
2	Qwen3.8-Flash-Next	125B MoE	Qwen Custom	community57 tok/s	47/50
3	DeepSeek V4.1 Flash	552B MoE	MIT	estimated29 tok/s	50/50
4	KAT-Coder-V2.5	35B MoE	Apache 2.0	estimated120 tok/s	39/50
5	MiniMax-M3	428B MoE	MiniMax Community	estimated42 tok/s	48/50
6	Qwen 3.6-35B-A3B	35B MoE	Apache 2.0	community153 tok/s	38/50
7	Muse Spark Open	816B MoE	Muse Model License	estimated31 tok/s	49/50
8	GLM 5.2	753B MoE	MIT	estimated22 tok/s	50/50

Every Mac in this tier (512 GB): Mac Studio M5 Ultra (512 GB) · Mac Studio M3 Ultra (512 GB)

Provenance key

sourced

Published figure — vendor or press review

estimated

Derived from published bandwidth model

community

Linked individual run — blog, forum or submission

🛒 Shopping by memory?

Memory is the one spec you can’t upgrade later. A Mac mini M6 (16 GB) runs 32 of the index’s models; with 32 GB it runs 54, and a 128 GB Mac Studio M5 Max runs 80.

Mac mini M6 on Amazon → Mac Studio M5 Max on Amazon → Compare all Macs →

As an Amazon Associate, LLMCheck earns from qualifying purchases. Affiliate links cost you nothing extra and never influence our rankings.

FAQ: which LLM fits my Mac’s RAM?

What is the best local LLM for a 8 GB Mac?

On a MacBook Air M1 (8 GB), the LLMCheck index's top pick is Nanbeige4.2-3B at ~30 tok/s (estimated). 21 models fit in 8 GB; the fastest is Gemma 4 E2B (~44 tok/s (estimated)) and the most capable is Bonsai 27B (no published figure on this chip).

What is the best local LLM for a 16 GB Mac?

On a Mac mini M6 (16 GB), the LLMCheck index's top pick is Gemma 4 12B at ~17 tok/s (estimated). 32 models fit in 16 GB; the fastest is Gemma 4 E2B (~91 tok/s (estimated)) and the most capable is Ternary Bonsai 2 27B (no published figure on this chip).

What is the best local LLM for a 24 GB Mac?

On a Mac mini M6 (24 GB), the LLMCheck index's top pick is Qwen 3.6-35B-A3B at 64 tok/s (community run). 54 models fit in 24 GB; the fastest is Gemma 4 E2B (~99 tok/s (estimated)) and the most capable is Qwen3.8-27B (9 tok/s (community run)).

What is the best local LLM for a 32 GB Mac?

On a Mac mini M6 (32 GB), the LLMCheck index's top pick is Qwen 3.6-35B-A3B at 64 tok/s (community run). 54 models fit in 32 GB; the fastest is Gemma 4 E2B (~99 tok/s (estimated)) and the most capable is Qwen3.8-27B (9 tok/s (community run)).

What is the best local LLM for a 64 GB Mac?

On a Mac mini M5 Pro (64 GB), the LLMCheck index's top pick is Qwen 3.6-35B-A3B at ~92 tok/s (estimated). 66 models fit in 64 GB; the fastest is Maple Preview 20B-A1B (387 tok/s (published figure)) and the most capable is Qwen3.8-27B (~15 tok/s (estimated)).

What is the best local LLM for a 128 GB Mac?

On a Mac Studio M5 Max (128 GB), the LLMCheck index's top pick is Qwen3.8-Flash-Next at 44 tok/s (community run). 80 models fit in 128 GB; the fastest is Gemma 4 E2B (~261 tok/s (estimated)) and the most capable is GLM-5.3-Flash (~27 tok/s (estimated)).

What is the best local LLM for a 192–256 GB Mac?

On a Mac Studio M5 Ultra (256 GB), the LLMCheck index's top pick is MiMo-V2.6-Flash at ~79 tok/s (estimated). 86 models fit in 192 GB, the smallest Mac in this tier; the fastest is Gemma 4 E2B (~378 tok/s (estimated)) and the most capable is MiniMax-M3 (~42 tok/s (estimated)).

What is the best local LLM for a 512 GB Mac?

On a Mac Studio M5 Ultra (512 GB), the LLMCheck index's top pick is MiMo-V2.6-Flash at ~79 tok/s (estimated). 104 models fit in 512 GB; the fastest is Gemma 4 E2B (~378 tok/s (estimated)) and the most capable is DeepSeek V4.1 Flash (~29 tok/s (estimated)).