Largest models with a runnable quantization

Frontier-scale open-weight models whose smallest quantization still lands within reach of a high-end workstation or a large unified-memory machine.

From the file· live filter over real dataFrom the file· 40 models

How this is ranked

Ranked by parameter count among models whose smallest published quantization is under 100 GiB — the practical ceiling for a single machine today.

Largest models with a runnable quantization

#	Model	Params○	Smallest quant●	Parameters○
1	MiniMax-M3MoE	427B	84.31 GiB	427B
2	Hermes-4-405B	406B	81.23 GiB	406B
3	Hermes-3-Llama-3.1-405B	406B	87.08 GiB	406B
4	Qwen3.5-397B-A17BMoE	403B	82.64 GiB	403B
5	Llama-4-Maverick-17B-128E-InstructMoE	402B	92.59 GiB	402B
6	Trinity-Large-ThinkingMoE	399B	75.64 GiB	399B
7	Trinity-Large-PreviewMoE	399B	76.06 GiB	399B
8	Trinity-Large-TrueBaseMoE	399B	73.49 GiB	399B
9	Ornith-1.0-397BMoE	397B	76.16 GiB	397B
10	Nex-N2-ProMoE	397B	76.16 GiB	397B
11	Qwen3-Coder-REAP-363B-A35BMoE	363B	94.37 GiB	363B
12	GLM-4.7MoE	358B	71.50 GiB	358B
13	GLM-4.5MoE	358B	71.81 GiB	358B
14	GLM-4.6MoE	357B	71.07 GiB	357B
15	GLM-4.6-Derestricted-v3MoE	357B	71.07 GiB	357B
16	MiMo-V2.5MoE	311B	60.09 GiB	311B
17	MiMo-V2-FlashMoE	310B	58.20 GiB	310B
18	DeepSeek-V4-Flash-0731MoE	304B	76.87 GiB	304B
19	ERNIE-4.5-300B-A47B-PT	300B	71.48 GiB	300B
20	Hy3MoE	299B	59.65 GiB	299B
21	DeepSeek-V4-FlashMoE	291B	76.87 GiB	291B
22	DeepSeek-V4-Flash-Q8_0-GGUF	284B	96.19 GiB	284B
23	grok-2MoE	270B	57.16 GiB	270B
24	GLM-4.6-REAP-268B-A32BMoE	269B	60.36 GiB	269B
25	Qwen3.5-REAP-262B-A17BMoE	262B	64.50 GiB	262B
26	Solar-Open2-250BMoE	250B	52.67 GiB	250B
27	DeepSeek-V2.5MoE	236B	44.14 GiB	236B
28	DeepSeek-Coder-V2-Instruct-0724MoE	236B	44.14 GiB	236B
29	DeepSeek-Coder-V2-InstructMoE	236B	44.14 GiB	236B
30	Qwen3-VL-235B-A22B-InstructMoE	236B	50.67 GiB	236B
31	Qwen3-VL-235B-A22B-ThinkingMoE	236B	50.83 GiB	236B
32	Qwen3-235B-A22BMoE	235B	79.81 GiB	235B
33	Qwen3-235B-A22B-Instruct-2507MoE	235B	49.57 GiB	235B
34	Qwen3-235B-A22B-Thinking-2507MoE	235B	49.57 GiB	235B
35	Qwen3-235B-A22B-abliteratedMoE	235B	44.66 GiB	235B
36	MiniMax-M2.7MoE	229B	43.47 GiB	229B
37	MiniMax-M2.5MoE	229B	43.32 GiB	229B
38	MiniMax-M2MoE	229B	43.78 GiB	229B
39	MiniMax-M2.1MoE	229B	43.32 GiB	229B
40	MiniMax-M2.7-BF16-ultra-uncensored-hereticMoE	229B	43.32 GiB	229B

○Spec sheet●From the file○Spec sheetwhat these mean

Best local models for coding Best local reasoning models Best models for an 8GB card Best models for long context Best permissively licensed models Best models under 4GB Best local speech recognition models Best local text-to-speech models Best local vision-language models Best local image generation models