DGX Spark vs. Framework Desktop for local AI

Two 128 GB unified-memory boxes with similar bandwidth (273 vs. 256 GB/s) at very different prices.

Total: $6,950
Usable memory: 96.0 GB
gpt-oss-120b Q4: Fits
Est. tok/s: up to 71

Parts

Part	Item	Price	Buy
GPU	NVIDIA DGX Spark 128GB	$6,950NVIDIA list price (not re-verified 2026-11-05: NVIDIA marketplace blocked/empty response; last confirmed $6,950 on marketplace listing; widely reported sold out in the US)	Amazon
Total		$6,950

We may earn a commission when you buy through these links, at no extra cost to you. Prices are dated snapshots (2026-10-23 for parts; GPUs at list price unless marked used). Used prices come from one dated listing each.

Will gpt-oss-120b run?

Model	Q4_K_M	Q8_0	FP16	Est. tok/s (Q4)
gpt-oss-120b82.5 GB at Q4 · MoE	Fits	Won’t fit	Won’t fit	up to 71

Tok/s = memory bandwidth ÷ model weight size × a 45–80% efficiency band, calibrated on llama.cpp scoreboards. Mixture-of-experts models show only an upper bound ("up to"). How we estimate speed.

Fit checks

NVIDIA DGX Spark 128GB is a complete system: no PSU, case or board to pick.

Compared with: Framework Desktop (Ryzen AI Max+ 395) 128GB

Total: $3,449
Usable memory: 96.0 GB
gpt-oss-120b Q4: Fits
Est. tok/s: up to 66

Part	Item	Price	Buy
GPU	Framework Desktop (Ryzen AI Max+ 395) 128GB	$3,449Framework list price, DIY Edition 128GB (pre-order, ships within 5 business days)	Amazon
Total		$3,449

Model	Q4_K_M	Q8_0	FP16	Est. tok/s (Q4)
gpt-oss-120b82.5 GB at Q4 · MoE	Fits	Won’t fit	Won’t fit	up to 66

All models on this build

Model	Q4_K_M	Q8_0	FP16	Est. tok/s (Q4)
Kimi-K3.1-Pro1939.3 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Kimi-K31939.1 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Qwen3.8-2.4T-A95B1706.4 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
LongCat-2.51241.8 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
LongCat-2.01238.9 GB at Q4	Won’t fit	Won’t fit	Won’t fit	—
DeepSeek-V4-Pro-08131151.7 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
DeepSeek-V4-Pro1115.7 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Kimi-K2.5716.9 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Kimi-K2.6716.9 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Kimi-K2.7-Code716.9 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Kimi-K2-Thinking716.6 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Kimi-K2-Instruct-0905716.6 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Kimi-K2-Instruct716.6 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
MiMo-V2.6-Pro-RL715.1 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
MiMo-V2.6-Pro-MOPD715.1 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
MiMo-V2.5-Pro714.4 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Ring-2.5-1T706.7 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Ling-2.5-1T706.7 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Ling-1T698.0 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Ring-1T698.0 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Inkling665.0 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Intern-S1-Pro640.4 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Muse-Spark-Open-816B-A42B574.1 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Hy4-preview544.8 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
GLM-5526.6 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
GLM-5.1526.6 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
GLM-5.3526.2 GB at Q4	Won’t fit	Won’t fit	Won’t fit	—
GLM-5.2526.2 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
GLM-5.3-BF16526.2 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
K-EXAONE-2.0-750B-A37B523.4 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
DeepSeek-V3.2-Exp478.9 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
DeepSeek-V3.2-Speciale478.9 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
DeepSeek-V3.2478.8 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
DeepSeek-V3-0324478.2 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
DeepSeek-R1-0528478.2 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
DeepSeek-V3.1-Base478.2 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
DeepSeek-R1478.2 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
DeepSeek-R1-Zero478.2 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Step-5-Preview422.3 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Ling-3.1-flash399.9 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Ling-3.1-flash-1019399.9 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
LongCat-Flash-Chat392.7 GB at Q4	Won’t fit	Won’t fit	Won’t fit	—
LongCat-Flash-Thinking392.7 GB at Q4	Won’t fit	Won’t fit	Won’t fit	—
LongCat-Flash-Thinking-2601392.7 GB at Q4	Won’t fit	Won’t fit	Won’t fit	—
NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16391.8 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Beam-1.0-501B-A23B350.5 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Qwen3-Coder-480B-A35B-Instruct335.8 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
MiniMax-M1-40k319.0 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
MiniMax-M1-80k319.0 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
MiniMax-M3298.7 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
ERNIE-4.5-VL-424B-A47B-PT296.3 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Intern-S2-397B282.3 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Qwen3.5-397B-A17B282.2 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Llama-4-Maverick-17B-128E-Instruct281.0 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Trinity-Large-Thinking278.9 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Trinity-Large-Preview278.9 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Trinity-Large-TrueBase278.9 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Trinity-Large-Base278.9 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Nex-N2.5-Pro277.6 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Nex-N2-Pro277.6 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
NorthCUA-Pro277.6 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
GLM-4.5250.8 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
GLM-4.7250.8 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
GLM-4.6249.7 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
GLM-5.3-Flash225.0 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Motif-3220.5 GB at Q4	Won’t fit	Won’t fit	Won’t fit	—
MiMo-V2.5217.7 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
MiMo-V2.6-Flash-RL217.7 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
MiMo-V2.6-Flash-MOPD217.7 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
MiMo-V2-Flash217.0 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
DeepSeek-V4-Flash-Vision-Exp213.4 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
DeepSeek-V4-Flash-0731213.1 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
ERNIE-4.5-300B-A47B-PT210.5 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Hy3209.3 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Hy3-preview209.3 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Ling-plus205.0 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
DeepSeek-V4-Flash203.8 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
MemT-1.0-Flash201.8 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
ERNIE-5.5-Flash-Base200.6 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Inkling-Small186.4 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Llama-3_1-Nemotron-Ultra-253B-v1177.7 GB at Q4	Won’t fit	Won’t fit	Won’t fit	—
Solar-Open2-250B175.5 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Intern-S1168.8 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
K-EXAONE-236B-A23B166.3 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
DeepSeek-V2.5165.4 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
DeepSeek-Coder-V2-Instruct165.4 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
DeepSeek-V2.5-1210165.4 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
DeepSeek-V2-Chat-0628165.4 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
DeepSeek-Coder-V2-Instruct-0724165.4 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Qwen3-VL-235B-A22B-Instruct165.3 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Qwen3-235B-A22B164.9 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Qwen3-235B-A22B-Instruct-2507164.9 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Qwen3-235B-A22B-Thinking-2507164.9 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
MiniMax-M2160.4 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
MiniMax-M2.5160.4 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
MiniMax-M2.7160.4 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
MiniMax-M2.1160.4 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
command-a-plus-05-2026-bf16153.5 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Step-3.7-Flash141.4 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Step-3.5-Flash140.0 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Step-3.5-Flash-Base-Midtrain138.9 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Qwen3.8-Flash-Next126.5 GB at Q4	Won’t fit	Won’t fit	Won’t fit	—
Mixtral-8x22B-Instruct-v0.199.0 GB at Q4 · MoE	Won’t fit	Won’t fit	Won’t fit	—
Ling-3.0-flash-Sante89.9 GB at Q4 · MoE	Tight	Won’t fit	Won’t fit	up to 71
Ling-3.0-flash89.9 GB at Q4 · MoE	Tight	Won’t fit	Won’t fit	up to 71
Ling-3.0-flash-Fin89.9 GB at Q4 · MoE	Tight	Won’t fit	Won’t fit	up to 71
Ornith-1.5-122B88.2 GB at Q4	Tight	Won’t fit	Won’t fit	1.6–2.9
Qwen3.5-122B-A10B88.2 GB at Q4 · MoE	Tight	Won’t fit	Won’t fit	up to 36
Devstral-2-123B-Instruct-251288.2 GB at Q4	Tight	Won’t fit	Won’t fit	1.6–2.9
Ling-3.0-flash-VL88.0 GB at Q4 · MoE	Tight	Won’t fit	Won’t fit	up to 65
NVIDIA-Nemotron-3-Super-120B-A12B-BF1687.2 GB at Q4 · MoE	Tight	Won’t fit	Won’t fit	up to 30
Mistral-Small-4-119B-260384.2 GB at Q4 · MoE	Fits	Won’t fit	Won’t fit	up to 55
Youtu-LLM-A7B-Base83.5 GB at Q4 · MoE	Fits	Won’t fit	Won’t fit	up to 51
Laguna-S-2.183.0 GB at Q4 · MoE	Fits	Won’t fit	Won’t fit	up to 45
gpt-oss-120b82.5 GB at Q4 · MoE	Fits	Won’t fit	Won’t fit	up to 71
c4ai-command-a-03-202578.4 GB at Q4	Fits	Won’t fit	Won’t fit	1.8–3.2
GLM-4.5-Air78.0 GB at Q4 · MoE	Fits	Won’t fit	Won’t fit	up to 30
Llama-4-Scout-17B-16E-Instruct76.7 GB at Q4 · MoE	Fits	Won’t fit	Won’t fit	up to 21
Llama-4-Scout-17B-16E76.7 GB at Q4 · MoE	Fits	Won’t fit	Won’t fit	up to 21
GLM-4.5V76.1 GB at Q4 · MoE	Fits	Won’t fit	Won’t fit	up to 30
GLM-4.6V76.1 GB at Q4 · MoE	Fits	Won’t fit	Won’t fit	up to 30
Ling-2.6-flash75.9 GB at Q4 · MoE	Fits	Won’t fit	Won’t fit	up to 49
sarvam-105b74.9 GB at Q4 · MoE	Fits	Won’t fit	Won’t fit	up to 35
Ling-flash-2.072.7 GB at Q4 · MoE	Fits	Won’t fit	Won’t fit	up to 59
LLaDA2.1-flash72.7 GB at Q4 · MoE	Fits	Won’t fit	Won’t fit	up to 59
Ring-flash-2.072.7 GB at Q4 · MoE	Fits	Won’t fit	Won’t fit	up to 59
LLaDA2.2-flash72.7 GB at Q4 · MoE	Fits	Won’t fit	Won’t fit	up to 59
Solar-Open-100B72.6 GB at Q4 · MoE	Fits	Won’t fit	Won’t fit	up to 30
Qwen3-Next-80B-A3B-Instruct57.7 GB at Q4 · MoE	Fits	Won’t fit	Won’t fit	up to 120
Qwen3-Next-80B-A3B-Thinking57.7 GB at Q4 · MoE	Fits	Won’t fit	Won’t fit	up to 120
AliceAI-Foundation-80B-A3B-Base57.7 GB at Q4 · MoE	Fits	Won’t fit	Won’t fit	up to 120
Hunyuan-A13B-Instruct57.0 GB at Q4 · MoE	Fits	Won’t fit	Won’t fit	up to 28
Qwen3-Coder-Next56.5 GB at Q4 · MoE	Fits	Won’t fit	Won’t fit	up to 120
Kolibri-155.5 GB at Q4	Fits	Won’t fit	Won’t fit	2.6–4.6
Qwen2.5-VL-72B-Instruct52.2 GB at Q4	Fits	Tight	Won’t fit	2.8–4.9
UI-TARS-72B-DPO52.2 GB at Q4	Fits	Tight	Won’t fit	2.8–4.9
Qwen2.5-72B-Instruct51.7 GB at Q4	Fits	Tight	Won’t fit	2.8–5.0
Apertus-70B-Instruct-250950.2 GB at Q4	Fits	Tight	Won’t fit	2.9–5.1
Apertus-70B-250950.2 GB at Q4	Fits	Tight	Won’t fit	2.9–5.1
Llama-3.3-70B-Instruct50.2 GB at Q4	Fits	Tight	Won’t fit	2.9–5.1
DeepSeek-R1-Distill-Llama-70B50.2 GB at Q4	Fits	Tight	Won’t fit	2.9–5.1
Llama-3.1-70B50.2 GB at Q4	Fits	Tight	Won’t fit	2.9–5.1
Hermes-2-Theta-Llama-3-70B50.2 GB at Q4	Fits	Tight	Won’t fit	2.9–5.1
Meta-Llama-3.1-70B-Instruct50.2 GB at Q4	Fits	Tight	Won’t fit	2.9–5.1
LongCat-Flash-Lite49.2 GB at Q4	Fits	Fits	Won’t fit	2.9–5.2
AI21-Jamba2-Mini37.0 GB at Q4 · MoE	Fits	Fits	Won’t fit	up to 30
Llama-3_1-Nemotron-51B-Instruct36.9 GB at Q4	Fits	Fits	Won’t fit	3.9–7.0
Llama-3_3-Nemotron-Super-49B-v1_535.8 GB at Q4	Fits	Fits	Won’t fit	4.1–7.2
Llama-3_3-Nemotron-Super-49B-v135.8 GB at Q4	Fits	Fits	Won’t fit	4.1–7.2
Phi-3.5-MoE-instruct30.2 GB at Q4 · MoE	Fits	Fits	Won’t fit	up to 55
GRIN-MoE30.2 GB at Q4 · MoE	Fits	Fits	Won’t fit	up to 55
ALIA-40b29.2 GB at Q4	Fits	Fits	Tight	5.0–8.9
IQuest-Coder-V1-40B-Loop-Instruct28.7 GB at Q4	Fits	Fits	Tight	5.1–9.1
IQuest-Coder-V1-40B-Instruct28.7 GB at Q4	Fits	Fits	Tight	5.1–9.1
Recursive Flint28.7 GB at Q4 · MoE	Fits	Fits	Tight	up to 144
K2-Horizon-MoVA-36B-A4B27.1 GB at Q4 · MoE	Fits	Fits	Tight	up to 90
Seed-OSS-36B-Instruct26.2 GB at Q4	Fits	Fits	Fits	5.6–10.0
Seed-OSS-36B-Base-woSyn26.2 GB at Q4	Fits	Fits	Fits	5.6–10.0
Intern-S2-Preview26.2 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
Ornith-1.5-35B-A3B26.1 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
Sage-A1.526.1 GB at Q4	Fits	Fits	Fits	5.6–10
Qwen3.6-35B-A3B26.1 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
Qwen3.5-35B-A3B26.1 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
Apex-35B-V1.026.1 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
MiniCPM-V-4.7-35B-A3B25.5 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
MemT-1.0-Nano25.5 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
Agents-A125.5 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
Nex-N2.5-mini25.5 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
Nex-N2-mini25.5 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
Axion-1.0-35B25.5 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
NorthCUA-Mini25.5 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
Clover-35B-A3B-Preview25.2 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
Qwen-AgentWorld-35B-A3B25.2 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
KAT-Coder-V2.5-Dev25.2 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
EXAONE-4.5-33B24.9 GB at Q4	Fits	Fits	Fits	5.9–10
Falcon-H1-34B-Instruct24.5 GB at Q4	Fits	Fits	Fits	6.0–11
Qwen2.5-VL-32B-Instruct24.3 GB at Q4	Fits	Fits	Fits	6.1–11
Laguna-XS.224.3 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
Qwen3-VL-32B-Instruct24.3 GB at Q4	Fits	Fits	Fits	6.1–11
HyperCLOVAX-SEED-Think-32B24.2 GB at Q4	Fits	Fits	Fits	6.1–11
FlexOlmo-7x7B-1T24.2 GB at Q4	Fits	Fits	Fits	6.1–11
GLM-Z1-Rumination-32B-041424.1 GB at Q4	Fits	Fits	Fits	6.1–11
Nanbeige4.3-33B24.1 GB at Q4	Fits	Fits	Fits	6.1–11
Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF1624.0 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
Qwen2.5-Coder-32B-Instruct23.8 GB at Q4	Fits	Fits	Fits	6.2–11
DeepSeek-R1-Distill-Qwen-32B23.8 GB at Q4	Fits	Fits	Fits	6.2–11
Qwen2.5-32B-Instruct23.8 GB at Q4	Fits	Fits	Fits	6.2–11
QwQ-32B23.8 GB at Q4	Fits	Fits	Fits	6.2–11
QwQ-32B-Preview23.8 GB at Q4	Fits	Fits	Fits	6.2–11
Qwen2.5-32B23.8 GB at Q4	Fits	Fits	Fits	6.2–11
Qwen2.5-Coder-32B23.8 GB at Q4	Fits	Fits	Fits	6.2–11
Qwen3-32B23.8 GB at Q4	Fits	Fits	Fits	6.2–11
GLM-Z1-32B-041423.7 GB at Q4	Fits	Fits	Fits	6.2–11
GLM-4-32B-041423.7 GB at Q4	Fits	Fits	Fits	6.2–11
OLMo-2-0325-32B-Instruct23.5 GB at Q4	Fits	Fits	Fits	6.3–11
OLMo-2-0325-32B23.5 GB at Q4	Fits	Fits	Fits	6.3–11
Olmo-3-32B-Think23.5 GB at Q4	Fits	Fits	Fits	6.3–11
Olmo-3-1125-32B23.5 GB at Q4	Fits	Fits	Fits	6.3–11
Olmo-3.1-32B-Instruct23.5 GB at Q4	Fits	Fits	Fits	6.3–11
Olmo-3.1-32B-Think23.5 GB at Q4	Fits	Fits	Fits	6.3–11
Olmo-3.1-32B-Instruct-DPO23.5 GB at Q4	Fits	Fits	Fits	6.3–11
granite-4.0-h-small23.5 GB at Q4 · MoE	Fits	Fits	Fits	up to 40
sarvam-30b23.4 GB at Q4 · MoE	Fits	Fits	Fits	up to 150
EXAONE-4.0-32B23.3 GB at Q4	Fits	Fits	Fits	6.3–11
EXAONE-4.0.1-32B23.3 GB at Q4	Fits	Fits	Fits	6.3–11
EXAONE-3.5-32B-Instruct23.3 GB at Q4	Fits	Fits	Fits	6.3–11
EXAONE-Deep-32B23.3 GB at Q4	Fits	Fits	Fits	6.3–11
NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF1623.0 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
NVIDIA-Nemotron-3-Nano-30B-A3B-BF1623.0 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
Nemotron-Cascade-2-30B-A3B23.0 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
Gemma 4 31B22.8 GB at Q4	Fits	Fits	Fits	6.5–12
gemma-4-31B22.8 GB at Q4	Fits	Fits	Fits	6.5–12
GLM-4.7-Flash22.8 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
Xing4.0-29B-A4B22.8 GB at Q4 · MoE	Fits	Fits	Fits	up to 90
Xing4.0-29B-Intent22.8 GB at Q4 · MoE	Fits	Fits	Fits	up to 90
Qwen3-VL-30B-A3B-Instruct22.7 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
Qwen3-VL-30B-A3B-Thinking22.7 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
UI-Venus-1.5-30B-A3B22.7 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
kanana-2-30b-a3b-instruct-260122.4 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
kanana-2-30b-a3b-instruct22.4 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
kanana-2-30b-a3b-thinking-260122.4 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
Qwen3-30B-A3B22.3 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
Qwen3-Coder-30B-A3B-Instruct22.3 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
Qwen3-30B-A3B-Instruct-250722.3 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
Qwen3-30B-A3B-Thinking-250722.3 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
Tongyi-DeepResearch-30B-A3B22.3 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
Rubicon-Preview22.3 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
North-Mini-Code-1.022.3 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
Hy-Memory-30B-A3B22.0 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
Hy-MT2-30B-A3B22.0 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
granite-switch-4.1-30b21.9 GB at Q4	Fits	Fits	Fits	6.8–12
Muse-Glimmer-30B21.8 GB at Q4	Fits	Fits	Fits	6.8–12
ERNIE-4.5-VL-28B-A3B-Thinking21.7 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
ERNIE-4.5-VL-28B-A3B-PT21.5 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
granite-4.2-30b21.4 GB at Q4	Fits	Fits	Fits	6.9–12
granite-4.1-30b21.1 GB at Q4	Fits	Fits	Fits	7.0–12
Raon-1 mini preview21.1 GB at Q4 · MoE	Fits	Fits	Fits	up to 84
Qwen3.8-27B20.4 GB at Q4	Fits	Fits	Fits	7.3–13
Tess-4.5-27B20.4 GB at Q4	Fits	Fits	Fits	7.3–13
Qwen3.6-27B20.4 GB at Q4	Fits	Fits	Fits	7.3–13
Qwen3.5-27B20.4 GB at Q4	Fits	Fits	Fits	7.3–13
gemma-3-27b-it20.1 GB at Q4	Fits	Fits	Fits	7.4–13
Arete-27B20.1 GB at Q4	Fits	Fits	Fits	7.4–13
gemma-2-27b-it20.0 GB at Q4	Fits	Fits	Fits	7.4–13
gemma-2-27b20.0 GB at Q4	Fits	Fits	Fits	7.4–13
Qwen3.8-27B-Longbook27-v219.9 GB at Q4	Fits	Fits	Fits	7.5–13
medgemma-27b-text-it19.8 GB at Q4	Fits	Fits	Fits	7.5–13
Hemmingway-1.519.8 GB at Q4	Fits	Fits	Fits	7.5–13
Trinity-Mini19.2 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
diffusiongemma-26B-A4B-it19.0 GB at Q4 · MoE	Fits	Fits	Fits	up to 90
Gemma 4 26B A4B19.0 GB at Q4 · MoE	Fits	Fits	Fits	up to 95
gemma-4-26B-A4B19.0 GB at Q4 · MoE	Fits	Fits	Fits	up to 90
Voxtral-Small-24B-250717.9 GB at Q4	Fits	Fits	Fits	8.4–15
Mistral-Small-3.2-24B-Instruct-250617.7 GB at Q4	Fits	Fits	Fits	8.4–15
Mistral-Small-3.1-24B-Instruct-250317.7 GB at Q4	Fits	Fits	Fits	8.4–15
LFM2.5-24B-A2B17.7 GB at Q4 · MoE	Fits	Fits	Fits	up to 180
LFM2-24B-A2B17.6 GB at Q4 · MoE	Fits	Fits	Fits	up to 180
Mistral-Small-24B-Instruct-250117.4 GB at Q4	Fits	Fits	Fits	8.6–15
EuroLLM-22B-Instruct-251216.8 GB at Q4	Fits	Fits	Fits	9.0–16
EuroLLM-22B-251216.8 GB at Q4	Fits	Fits	Fits	9.0–16
solar-pro-preview-instruct16.4 GB at Q4	Fits	Fits	Fits	9.2–16
ERNIE-4.5-21B-A3B-PT16.3 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
ERNIE-4.5-21B-A3B-Thinking16.2 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
gpt-oss-20b15.6 GB at Q4 · MoE	Fits	Fits	Fits	up to 100
reka-flash-315.6 GB at Q4	Fits	Fits	Fits	9.7–17
maple-preview15.1 GB at Q4 · MoE	Fits	Fits	Fits	up to 360
internlm2_5-20b-chat14.8 GB at Q4	Fits	Fits	Fits	10–18
Groot-19B-A3B-26102314.5 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
Ling-Coder-lite12.7 GB at Q4 · MoE	Fits	Fits	Fits	up to 131
Ring-lite-distill-preview12.7 GB at Q4 · MoE	Fits	Fits	Fits	up to 131
Ling-lite12.7 GB at Q4 · MoE	Fits	Fits	Fits	up to 131
Ling-lite-1.5-250712.7 GB at Q4 · MoE	Fits	Fits	Fits	up to 131
Ling-lite-1.512.7 GB at Q4 · MoE	Fits	Fits	Fits	up to 131
Ring-lite12.7 GB at Q4 · MoE	Fits	Fits	Fits	up to 131
Ring-mini-linear-2.012.5 GB at Q4 · MoE	Fits	Fits	Fits	up to 225
Kimi-VL-A3B-Instruct12.4 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
Kimi-VL-A3B-Thinking12.4 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
Kimi-VL-A3B-Thinking-250612.4 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
LLaDA2.1-mini12.3 GB at Q4 · MoE	Fits	Fits	Fits	up to 257
LLaDA2.0-mini12.3 GB at Q4 · MoE	Fits	Fits	Fits	up to 257
Ling-mini-2.012.3 GB at Q4 · MoE	Fits	Fits	Fits	up to 257
Ring-mini-2.012.3 GB at Q4 · MoE	Fits	Fits	Fits	up to 257
LLaDA2.2-mini12.3 GB at Q4 · MoE	Fits	Fits	Fits	up to 257
LLaDA2.0-mini-preview12.3 GB at Q4 · MoE	Fits	Fits	Fits	up to 257
Moonlight-16B-A3B-Instruct12.1 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
Moonlight-16B-A3B12.1 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
Instella-MoE-16B-A3B-Think12.1 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
DeepSeek-Coder-V2-Lite-Instruct12.0 GB at Q4 · MoE	Fits	Fits	Fits	up to 150
kanana-1.5-15.7b-a3b-instruct11.9 GB at Q4 · MoE	Fits	Fits	Fits	up to 120
Phi-4-reasoning-vision-15B11.5 GB at Q4	Fits	Fits	Fits	13–24
Apriel-Nemotron-15b-Thinker11.4 GB at Q4	Fits	Fits	Fits	14–24
Apriel-1.5-15b-Thinker11.4 GB at Q4	Fits	Fits	Fits	14–24
Apriel-1.6-15b-Thinker11.4 GB at Q4	Fits	Fits	Fits	14–24
Carbon-X (DNA)11.3 GB at Q4	Fits	Fits	Fits	14–24
Qwen2.5-Coder-14B-Instruct11.3 GB at Q4	Fits	Fits	Fits	14–24
DeepSeek-R1-Distill-Qwen-14B11.3 GB at Q4	Fits	Fits	Fits	14–24
Qwen2.5-14B-Instruct11.3 GB at Q4	Fits	Fits	Fits	14–24
Qwen2.5-14B11.3 GB at Q4	Fits	Fits	Fits	14–24
Qwen2.5-14B-Instruct-1M11.3 GB at Q4	Fits	Fits	Fits	14–24
Qwen2.5-Coder-14B11.3 GB at Q4	Fits	Fits	Fits	14–24
Qwen3-14B11.3 GB at Q4	Fits	Fits	Fits	14–24
JanusCoder-14B11.3 GB at Q4	Fits	Fits	Fits	14–24
HyperCLOVAX-SEED-Think-14B11.3 GB at Q4	Fits	Fits	Fits	14–24
phi-411.2 GB at Q4	Fits	Fits	Fits	14–25
Phi-4-reasoning11.2 GB at Q4	Fits	Fits	Fits	14–25
Phi-4-reasoning-plus11.2 GB at Q4	Fits	Fits	Fits	14–25
OLMo-2-1124-13B-Instruct10.6 GB at Q4	Fits	Fits	Fits	15–26
Mistral-Nemo-Instruct-24079.5 GB at Q4	Fits	Fits	Fits	17–29
gemma-3-12b-it9.5 GB at Q4	Fits	Fits	Fits	17–30
gemma-3-12b-pt9.5 GB at Q4	Fits	Fits	Fits	17–30
Mellum2.1-12B-A2.5B-Thinking9.5 GB at Q4 · MoE	Fits	Fits	Fits	up to 144
Mellum2-12B-A2.5B-Thinking9.5 GB at Q4 · MoE	Fits	Fits	Fits	up to 144
Mellum2-12B-A2.5B-Instruct9.5 GB at Q4 · MoE	Fits	Fits	Fits	up to 144
Mellum2-12B-A2.5B-Thinking-SFT9.5 GB at Q4 · MoE	Fits	Fits	Fits	up to 144
Mellum2.1-12B-A2.5B-Thinking-MTP9.5 GB at Q4 · MoE	Fits	Fits	Fits	up to 144
Gemma 4 12B9.3 GB at Q4	Fits	Fits	Fits	17–30
HyperCLOVAX-SEED-Omni-8B8.5 GB at Q4	Fits	Fits	Fits	19–34
Llama-3.2-11B-Vision-Instruct8.4 GB at Q4	Fits	Fits	Fits	19–34
Llama-3.2-11B-Vision8.4 GB at Q4	Fits	Fits	Fits	19–34
Falcon3-10B-Instruct8.2 GB at Q4	Fits	Fits	Fits	20–35
GLM-4.1V-9B-Thinking8.2 GB at Q4	Fits	Fits	Fits	20–35
GLM-4.6V-Flash8.2 GB at Q4	Fits	Fits	Fits	20–35
GLM-4.1V-9B-Base8.2 GB at Q4	Fits	Fits	Fits	20–35
Glyph8.2 GB at Q4	Fits	Fits	Fits	20–35
INTELLECT-1-Instruct8.1 GB at Q4	Fits	Fits	Fits	20–35
INTELLECT-18.1 GB at Q4	Fits	Fits	Fits	20–35
ZAYA1-VL-8B7.8 GB at Q4	Fits	Fits	Fits	21–37
SenseNova-Loop-9B-Preview7.7 GB at Q4	Fits	Fits	Fits	21–37
Ornith-1.5-9B7.7 GB at Q4	Fits	Fits	Fits	21–37
Qwen3.5-9B7.7 GB at Q4	Fits	Fits	Fits	21–37
MiniCPM-SALA7.6 GB at Q4	Fits	Fits	Fits	21–38
NorthCUA-Nano7.6 GB at Q4	Fits	Fits	Fits	22–38
GLM-4-9B-04147.6 GB at Q4	Fits	Fits	Fits	22–38
GLM-Z1-9B-04147.6 GB at Q4	Fits	Fits	Fits	22–38
codegeex4-all-9b7.6 GB at Q4	Fits	Fits	Fits	22–38
glm-4-9b-chat-hf7.6 GB at Q4	Fits	Fits	Fits	22–38
moondream3-preview7.5 GB at Q4	Fits	Fits	Fits	22–39
gemma-2-9b-it7.4 GB at Q4	Fits	Fits	Fits	22–39
gemma-2-9b7.4 GB at Q4	Fits	Fits	Fits	22–39
Gemma-SEA-LION-v3-9B-IT7.4 GB at Q4	Fits	Fits	Fits	22–39
EuroLLM-9B-Instruct-25127.4 GB at Q4	Fits	Fits	Fits	22–39
NVIDIA-Nemotron-Nano-9B-v27.2 GB at Q4	Fits	Fits	Fits	23–41
ZAYA1-base7.2 GB at Q4	Fits	Fits	Fits	23–41
Yi-Coder-9B-Chat7.2 GB at Q4	Fits	Fits	Fits	23–41
Yi-Coder-9B7.2 GB at Q4	Fits	Fits	Fits	23–41
internlm3-8b-instruct7.1 GB at Q4	Fits	Fits	Fits	23–41
granite-4.1-8b7.1 GB at Q4	Fits	Fits	Fits	23–41
granite-4.2-8b7.1 GB at Q4	Fits	Fits	Fits	23–41
granite-switch-4.1-8b7.1 GB at Q4	Fits	Fits	Fits	23–41
Qwen3-VL-8B-Instruct7.1 GB at Q4	Fits	Fits	Fits	23–41
Qwen3-VL-8B-Thinking7.1 GB at Q4	Fits	Fits	Fits	23–41
UI-Venus-1.5-8B7.1 GB at Q4	Fits	Fits	Fits	23–41
Llama-3.1-Nemotron-Nano-VL-8B-V17.1 GB at Q4	Fits	Fits	Fits	23–41
MiniCPM-V-4_57.1 GB at Q4	Fits	Fits	Fits	23–41
aya-vision-8b7.0 GB at Q4	Fits	Fits	Fits	23–42
Intern-S1-mini7.0 GB at Q4	Fits	Fits	Fits	24–42
Nemotron-Labs-Diffusion-8B-Base6.9 GB at Q4	Fits	Fits	Fits	24–42
LFM2.5-8B-A1B6.9 GB at Q4 · MoE	Fits	Fits	Fits	up to 360
Mistral-NeMo-Minitron-8B-Base6.9 GB at Q4	Fits	Fits	Fits	24–43
LFM2-8B-A1B6.8 GB at Q4 · MoE	Fits	Fits	Fits	up to 360
rnj-1-instruct6.8 GB at Q4	Fits	Fits	Fits	24–43
MiMo-VL-7B-RL-25086.8 GB at Q4	Fits	Fits	Fits	24–43
MiMo-VL-7B-RL6.8 GB at Q4	Fits	Fits	Fits	24–43
MiMo-Embodied-7B6.8 GB at Q4	Fits	Fits	Fits	24–43
MiMo-VL-7B-SFT6.8 GB at Q4	Fits	Fits	Fits	24–43
Qwen2.5-VL-7B-Instruct6.8 GB at Q4	Fits	Fits	Fits	24–43
UI-TARS-1.5-7B6.8 GB at Q4	Fits	Fits	Fits	24–43
Fara-7B6.8 GB at Q4	Fits	Fits	Fits	24–43
UI-Venus-Ground-7B6.8 GB at Q4	Fits	Fits	Fits	24–43
Qwen2-VL-7B-Instruct6.8 GB at Q4	Fits	Fits	Fits	24–43
UI-TARS-7B-DPO6.8 GB at Q4	Fits	Fits	Fits	24–43
UI-TARS-7B-SFT6.8 GB at Q4	Fits	Fits	Fits	24–43
Seed-Coder-8B-Instruct6.8 GB at Q4	Fits	Fits	Fits	25–44
Seed-Coder-8B-Reasoning6.8 GB at Q4	Fits	Fits	Fits	25–44
Stable-DiffCoder-8B-Instruct6.8 GB at Q4	Fits	Fits	Fits	25–44
Qwen3-8B6.7 GB at Q4	Fits	Fits	Fits	25–44
DeepSeek-R1-0528-Qwen3-8B6.7 GB at Q4	Fits	Fits	Fits	25–44
Qwen3-8B-Base6.7 GB at Q4	Fits	Fits	Fits	25–44
WeDLM-8B-Instruct6.7 GB at Q4	Fits	Fits	Fits	25–44
PaCoRe-8B6.7 GB at Q4	Fits	Fits	Fits	25–44
MiniCPM4-8B6.7 GB at Q4	Fits	Fits	Fits	25–44
MiniCPM4.1-8B6.7 GB at Q4	Fits	Fits	Fits	25–44
AgentCPM-Report6.7 GB at Q4	Fits	Fits	Fits	25–44
MiniCPM4-MCP6.7 GB at Q4	Fits	Fits	Fits	25–44
granite-3.3-8b-instruct6.7 GB at Q4	Fits	Fits	Fits	25–44
granite-3.1-8b-instruct6.7 GB at Q4	Fits	Fits	Fits	25–44
granite-3.2-8b-instruct6.7 GB at Q4	Fits	Fits	Fits	25–44
granite-3.0-8b-instruct6.7 GB at Q4	Fits	Fits	Fits	25–44
Apertus-8B-Instruct-25096.6 GB at Q4	Fits	Fits	Fits	25–45
Apertus-8B-25096.6 GB at Q4	Fits	Fits	Fits	25–45
Llama-3.1-Tulu-3-8B6.6 GB at Q4	Fits	Fits	Fits	25–45
Llama-3.1-Tulu-3-8B-DPO6.6 GB at Q4	Fits	Fits	Fits	25–45
Llama-3.1-Tulu-3.1-8B6.6 GB at Q4	Fits	Fits	Fits	25–45
kanana-1.5-8b-instruct-25056.6 GB at Q4	Fits	Fits	Fits	25–45
Llama-3.1-8B-Instruct6.6 GB at Q4	Fits	Fits	Fits	25–45
Llama-3.1-8B6.6 GB at Q4	Fits	Fits	Fits	25–45
DeepSeek-R1-Distill-Llama-8B6.6 GB at Q4	Fits	Fits	Fits	25–45
Meta-Llama-3.1-8B-Instruct6.6 GB at Q4	Fits	Fits	Fits	25–45
L3-8B-Stheno-v3.26.6 GB at Q4	Fits	Fits	Fits	25–45
Llama-3.1-Nemotron-Nano-8B-v16.6 GB at Q4	Fits	Fits	Fits	25–45
Meta-Llama-3.1-8B6.6 GB at Q4	Fits	Fits	Fits	25–45
LongWriter-llama3.1-8b6.6 GB at Q4	Fits	Fits	Fits	25–45
Hy-MT2-7B6.6 GB at Q4	Fits	Fits	Fits	25–45
Nemotron 3.5 Bolt E4B6.6 GB at Q4	Fits	Fits	Fits	25–45
Gemma 4 E4B6.6 GB at Q4	Fits	Fits	Fits	25–45
Ling-3.0-tiny6.5 GB at Q4 · MoE	Fits	Fits	Fits	up to 277
MiMo-7B-RL6.5 GB at Q4	Fits	Fits	Fits	26–46
MiMo-7B-Base6.5 GB at Q4	Fits	Fits	Fits	26–46
MiMo-7B-RL-05306.5 GB at Q4	Fits	Fits	Fits	26–46
MiMo-7B-SFT6.5 GB at Q4	Fits	Fits	Fits	26–46
EXAONE-3.5-7.8B-Instruct6.5 GB at Q4	Fits	Fits	Fits	26–46
EXAONE-Deep-7.8B6.5 GB at Q4	Fits	Fits	Fits	26–46
salamandra-7b-instruct6.4 GB at Q4	Fits	Fits	Fits	26–46
salamandra-7b6.4 GB at Q4	Fits	Fits	Fits	26–46
salamandra-7b-instruct-26066.4 GB at Q4	Fits	Fits	Fits	26–46
FastVLM-7B6.4 GB at Q4	Fits	Fits	Fits	26–46
Molmo2-O-7B6.4 GB at Q4	Fits	Fits	Fits	26–46
internlm2_5-7b-chat6.4 GB at Q4	Fits	Fits	Fits	26–47
internlm2_5-7b-chat-1m6.4 GB at Q4	Fits	Fits	Fits	26–47
Phi-mini-MoE-instruct6.3 GB at Q4 · MoE	Fits	Fits	Fits	up to 150
Qwen2.5-Coder-7B-Instruct6.3 GB at Q4	Fits	Fits	Fits	27–47
DeepSeek-R1-Distill-Qwen-7B6.3 GB at Q4	Fits	Fits	Fits	27–47
Qwen2.5-7B-Instruct6.3 GB at Q4	Fits	Fits	Fits	27–47
Qwen2.5-7B6.3 GB at Q4	Fits	Fits	Fits	27–47
Qwen2-7B-Instruct6.3 GB at Q4	Fits	Fits	Fits	27–47
Qwen2.5-Coder-7B6.3 GB at Q4	Fits	Fits	Fits	27–47
Qwen2-7B6.3 GB at Q4	Fits	Fits	Fits	27–47
Qwen2.5-7B-Instruct-1M6.3 GB at Q4	Fits	Fits	Fits	27–47
Qwen2.5-Math-7B6.3 GB at Q4	Fits	Fits	Fits	27–47
Qwen2.5-Math-7B-Instruct6.3 GB at Q4	Fits	Fits	Fits	27–47
Marco-o16.3 GB at Q4	Fits	Fits	Fits	27–47
SeaLLMs-v3-7B-Chat6.3 GB at Q4	Fits	Fits	Fits	27–47
Arcee-Spark6.3 GB at Q4	Fits	Fits	Fits	27–47
Falcon-H1-7B-Instruct6.3 GB at Q4	Fits	Fits	Fits	27–47
Falcon-H1R-7B6.3 GB at Q4	Fits	Fits	Fits	27–47
starvector-8b-im2svg6.2 GB at Q4	Fits	Fits	Fits	27–48
Falcon3-7B-Instruct6.2 GB at Q4	Fits	Fits	Fits	27–48
Olmo-Hybrid-7B6.2 GB at Q4	Fits	Fits	Fits	27–48
LLaDA-MoE-7B-A1B-Base6.1 GB at Q4 · MoE	Fits	Fits	Fits	up to 360
Zamba2-7B-Instruct6.1 GB at Q4	Fits	Fits	Fits	28–49
OLMo-2-1124-7B-Instruct6.1 GB at Q4	Fits	Fits	Fits	28–49
Olmo-3-7B-Think6.1 GB at Q4	Fits	Fits	Fits	28–49
Olmo-3-7B-Instruct6.1 GB at Q4	Fits	Fits	Fits	28–49
Olmo-3-1025-7B6.1 GB at Q4	Fits	Fits	Fits	28–49
Olmo-3-7B-Instruct-SFT6.1 GB at Q4	Fits	Fits	Fits	28–49
Olmo-3-7B-Think-SFT6.1 GB at Q4	Fits	Fits	Fits	28–49
Olmo-3-7B-Instruct-DPO6.1 GB at Q4	Fits	Fits	Fits	28–49
Olmo-3-7B-RL-Zero-Math6.1 GB at Q4	Fits	Fits	Fits	28–49
Olmo-3.1-7B-RL-Zero-Math6.1 GB at Q4	Fits	Fits	Fits	28–49
Olmo-3-7B-Think-DPO6.1 GB at Q4	Fits	Fits	Fits	28–49
Mistral-7B-Instruct-v0.36.1 GB at Q4	Fits	Fits	Fits	28–50
granite-4.0-h-tiny5.8 GB at Q4 · MoE	Fits	Fits	Fits	up to 360
OLMoE-1B-7B-0924-Instruct5.8 GB at Q4 · MoE	Fits	Fits	Fits	up to 277
OLMoE-1B-7B-09245.8 GB at Q4 · MoE	Fits	Fits	Fits	up to 277
OLMoE-1B-7B-0125-Instruct5.8 GB at Q4 · MoE	Fits	Fits	Fits	up to 277
OLMoE-1B-7B-01255.8 GB at Q4 · MoE	Fits	Fits	Fits	up to 277
OLMo-7B-0724-hf5.8 GB at Q4	Fits	Fits	Fits	29–52
granite-4.0-tiny-preview5.7 GB at Q4 · MoE	Fits	Fits	Fits	up to 360
YuE-s1-7B-anneal-en-cot5.3 GB at Q4	Fits	Fits	Fits	33–58
Trinity-Nano-Preview5.3 GB at Q4 · MoE	Fits	Fits	Fits	up to 360
Trinity-Nano-Base5.3 GB at Q4 · MoE	Fits	Fits	Fits	up to 360
gemma-3n-E2B-it4.8 GB at Q4	Fits	Fits	Fits	37–66
Youtu-VL-4B-Instruct4.7 GB at Q4	Fits	Fits	Fits	38–67
Gemma 4 E2B4.6 GB at Q4	Fits	Fits	Fits	40–70
Molmo2-ER4.4 GB at Q4	Fits	Fits	Fits	42–74
Apriel-5B-Instruct4.4 GB at Q4	Fits	Fits	Fits	42–75
WeVisOne-VL-4B4.4 GB at Q4	Fits	Fits	Fits	42–75
Mage-VL4.3 GB at Q4	Fits	Fits	Fits	43–76
Mage-VL-Cowork4.3 GB at Q4	Fits	Fits	Fits	43–76
Qwen3.5-4B4.2 GB at Q4	Fits	Fits	Fits	43–77
S5-nano4.2 GB at Q4	Fits	Fits	Fits	43–77
AFM-4.5B4.2 GB at Q4	Fits	Fits	Fits	44–78
Llama-3.1-Minitron-4B-Width-Base4.1 GB at Q4	Fits	Fits	Fits	45–80
Qwen3-VL-4B-Instruct4.1 GB at Q4	Fits	Fits	Fits	46–81
CapRL-Qwen3VL-4B4.1 GB at Q4	Fits	Fits	Fits	46–81
glm-edge-4b-chat4.0 GB at Q4	Fits	Fits	Fits	47–83
gemma-3-4b-it4.0 GB at Q4	Fits	Fits	Fits	47–84
medgemma-4b-it4.0 GB at Q4	Fits	Fits	Fits	47–84
medgemma-1.5-4b-it4.0 GB at Q4	Fits	Fits	Fits	47–84
gemma-3-4b-pt4.0 GB at Q4	Fits	Fits	Fits	47–84
Nanbeige4.2-3B3.9 GB at Q4	Fits	Fits	Fits	49–86
Phi-3.5-vision-instruct3.9 GB at Q4	Fits	Fits	Fits	49–87
Spark-X2.6-4B3.9 GB at Q4	Fits	Fits	Fits	49–88
Spark-X2.5-4B3.9 GB at Q4	Fits	Fits	Fits	49–88
POINTS-Reader3.9 GB at Q4	Fits	Fits	Fits	49–88
MiniCPM-V-43.8 GB at Q4	Fits	Fits	Fits	50–89
Qwen3-4B3.8 GB at Q4	Fits	Fits	Fits	50–90
Qwen3-4B-Instruct-25073.8 GB at Q4	Fits	Fits	Fits	50–90
Qwen3-4B-Thinking-25073.8 GB at Q4	Fits	Fits	Fits	50–90
Mellum-4b-base3.8 GB at Q4	Fits	Fits	Fits	50–90
granite-4.0-3b-vision3.8 GB at Q4	Fits	Fits	Fits	51–90
granite-vision-4.1-4b3.8 GB at Q4	Fits	Fits	Fits	51–90
Nemotron Labs Dory 1.0 4B3.8 GB at Q4	Fits	Fits	Fits	51–91
NVIDIA-Nemotron-3-Nano-4B-BF163.8 GB at Q4	Fits	Fits	Fits	51–91
Nanbeige4.1-3B3.7 GB at Q4	Fits	Fits	Fits	52–92
Nanbeige4-3B-Thinking-25113.7 GB at Q4	Fits	Fits	Fits	52–92
Phi-4-mini-flash-reasoning3.7 GB at Q4	Fits	Fits	Fits	53–94
Phi-4-mini-instruct3.7 GB at Q4	Fits	Fits	Fits	53–94
Phi-4-mini-reasoning3.7 GB at Q4	Fits	Fits	Fits	53–94
Nemotron-Labs-Diffusion-3B3.7 GB at Q4	Fits	Fits	Fits	53–94
Phi-3.5-mini-instruct3.7 GB at Q4	Fits	Fits	Fits	53–94
MediPhi-Instruct3.7 GB at Q4	Fits	Fits	Fits	53–94
MediPhi3.7 GB at Q4	Fits	Fits	Fits	53–94
MediPhi-Clinical3.7 GB at Q4	Fits	Fits	Fits	53–94
MediPhi-PubMed3.7 GB at Q4	Fits	Fits	Fits	53–94
HY-Embodied-0.53.6 GB at Q4	Fits	Fits	Fits	54–95
Phi-tiny-MoE-instruct3.6 GB at Q4 · MoE	Fits	Fits	Fits	up to 328
CapRL-3B3.6 GB at Q4	Fits	Fits	Fits	54–96
HyperCLOVAX-SEED-Vision-Instruct-3B3.6 GB at Q4	Fits	Fits	Fits	54–97
kanana-1.5-v-3b-instruct3.6 GB at Q4	Fits	Fits	Fits	55–98
granite-4.2-3b3.6 GB at Q4	Fits	Fits	Fits	55–98
kanana-2-3b-instruct3.4 GB at Q4	Fits	Fits	Fits	58–103
granite-4.1-3b3.4 GB at Q4	Fits	Fits	Fits	60–106
granite-4.0-micro3.4 GB at Q4	Fits	Fits	Fits	60–106
granite-3.1-3b-a800m-instruct3.3 GB at Q4 · MoE	Fits	Fits	Fits	up to 450
Falcon3-3B-Instruct3.3 GB at Q4	Fits	Fits	Fits	63–112
Llama-3.2-3B-Instruct3.2 GB at Q4	Fits	Fits	Fits	63–112
Llama-3.2-3B3.2 GB at Q4	Fits	Fits	Fits	63–112
Hermes-3-Llama-3.2-3B3.2 GB at Q4	Fits	Fits	Fits	63–112
DeepHermes-3-Llama-3-3B-Preview3.2 GB at Q4	Fits	Fits	Fits	63–112
AI21-Jamba-Reasoning-3B3.2 GB at Q4	Fits	Fits	Fits	63–113
granite-4.0-h-micro3.2 GB at Q4	Fits	Fits	Fits	63–113
Falcon-H1-3B-Instruct3.2 GB at Q4	Fits	Fits	Fits	64–114
d1-3B3.2 GB at Q4	Fits	Fits	Fits	65–115
LFM2.5-VL-3B3.2 GB at Q4	Fits	Fits	Fits	65–115
Qwen2.5-3B-Instruct3.2 GB at Q4	Fits	Fits	Fits	66–117
Qwen2.5-Coder-3B-Instruct3.2 GB at Q4	Fits	Fits	Fits	66–117
Qwen2.5-3B3.2 GB at Q4	Fits	Fits	Fits	66–117
Qwen2.5-Coder-3B3.2 GB at Q4	Fits	Fits	Fits	66–117
SmolLM3-3B3.1 GB at Q4	Fits	Fits	Fits	66–117
AI21-Jamba2-3B3.1 GB at Q4	Fits	Fits	Fits	67–119
granite-swash-3b-a600m3.1 GB at Q4 · MoE	Fits	Fits	Fits	up to 602
granite-vision-3.2-2b3.1 GB at Q4	Fits	Fits	Fits	68–121
granite-vision-3.1-2b-preview3.1 GB at Q4	Fits	Fits	Fits	68–121
granite-vision-3.3-2b-chart2csv-preview3.1 GB at Q4	Fits	Fits	Fits	68–121
LFM2.5-2.6B2.9 GB at Q4	Fits	Fits	Fits	75–134
Zamba2-2.7B2.9 GB at Q4	Fits	Fits	Fits	76–135
Zamba2-2.7B-instruct2.9 GB at Q4	Fits	Fits	Fits	76–135
granite-3.0-2b-instruct2.8 GB at Q4	Fits	Fits	Fits	77–137
gemma-2-2b-it2.8 GB at Q4	Fits	Fits	Fits	78–138
gemma-2-2b2.8 GB at Q4	Fits	Fits	Fits	78–138
EuroMoE-2.6B-A0.6B-Instruct-Preview2.8 GB at Q4 · MoE	Fits	Fits	Fits	up to 600
Motif-2.6B2.8 GB at Q4	Fits	Fits	Fits	78–139
LFM2-2.6B2.8 GB at Q4	Fits	Fits	Fits	79–140
LFM2-2.6B-Exp2.8 GB at Q4	Fits	Fits	Fits	79–140
LFM2-2.6B-Transcript2.8 GB at Q4	Fits	Fits	Fits	79–140
granite-3.3-2b-instruct2.8 GB at Q4	Fits	Fits	Fits	80–142
granite-3.1-2b-instruct2.8 GB at Q4	Fits	Fits	Fits	80–142
granite-3.2-2b-instruct2.8 GB at Q4	Fits	Fits	Fits	80–142
MiniCPM5-2B2.8 GB at Q4	Fits	Fits	Fits	81–143
MiniCPM5-2B-SFT2.8 GB at Q4	Fits	Fits	Fits	81–143
MiniCPM5-2B-Midtrain2.8 GB at Q4	Fits	Fits	Fits	81–143
sarvam-1-v0.52.7 GB at Q4	Fits	Fits	Fits	81–144
North-Micro-Vision-Instruct2.7 GB at Q4	Fits	Fits	Fits	82–145
UI-TARS-2B-SFT2.7 GB at Q4	Fits	Fits	Fits	83–147
WeVisOne-VL-2B2.7 GB at Q4	Fits	Fits	Fits	83–148
UI-Venus-1.5-2B2.7 GB at Q4	Fits	Fits	Fits	83–148
bitnet-b1.58-2B-4T-bf162.7 GB at Q4	Fits	Fits	Fits	84–149
EXAONE-3.5-2.4B-Instruct2.7 GB at Q4	Fits	Fits	Fits	84–150
EXAONE-Deep-2.4B2.7 GB at Q4	Fits	Fits	Fits	84–150
kanana-1.5-2.1b-instruct-25052.6 GB at Q4	Fits	Fits	Fits	87–155
Qwen3.5-2B2.6 GB at Q4	Fits	Fits	Fits	89–158
salamandra-2b2.6 GB at Q4	Fits	Fits	Fits	90–160
salamandra-2b-instruct2.6 GB at Q4	Fits	Fits	Fits	90–160
SmolVLM2-2.2B-Instruct2.6 GB at Q4	Fits	Fits	Fits	90–160
SmolVLM-Base2.6 GB at Q4	Fits	Fits	Fits	90–160
Qwen2-VL-2B-Instruct2.5 GB at Q4	Fits	Fits	Fits	92–163
LFM3-Omni-2B2.5 GB at Q4	Fits	Fits	Fits	94–166
granite-swash-2b2.5 GB at Q4	Fits	Fits	Fits	94–168
Qwen3-VL-2B-Instruct2.5 GB at Q4	Fits	Fits	Fits	95–169
kanana-nano-2.1b-instruct2.5 GB at Q4	Fits	Fits	Fits	97–173
Hy-MT2-1.8B2.4 GB at Q4	Fits	Fits	Fits	99–177
Qwen3-1.7B2.4 GB at Q4	Fits	Fits	Fits	100–177
Youtu-LLM-2B2.4 GB at Q4	Fits	Fits	Fits	103–184
LFM3-VL-2B2.3 GB at Q4	Fits	Fits	Fits	106–188
d1-2B2.3 GB at Q4	Fits	Fits	Fits	106–188
FastVLM-1.5B2.3 GB at Q4	Fits	Fits	Fits	106–189
internlm2_5-1_8b-chat2.3 GB at Q4	Fits	Fits	Fits	107–191
DeepSeek-R1-Distill-Qwen-1.5B2.2 GB at Q4	Fits	Fits	Fits	114–203
LFM3-1.7B2.2 GB at Q4	Fits	Fits	Fits	117–208
Qwen3-1.7B-Base2.2 GB at Q4	Fits	Fits	Fits	118–209
SmolLM2-1.7B2.2 GB at Q4	Fits	Fits	Fits	118–211
SmolLM-1.7B2.2 GB at Q4	Fits	Fits	Fits	118–211
Spark-X2.6-1.7B2.2 GB at Q4	Fits	Fits	Fits	119–211
Falcon3-1B-Instruct2.2 GB at Q4	Fits	Fits	Fits	121–216
EuroLLM-1.7B-Instruct2.2 GB at Q4	Fits	Fits	Fits	122–217
granite-4.0-1b2.1 GB at Q4	Fits	Fits	Fits	124–221
LFM2.5-VL-1.6B2.1 GB at Q4	Fits	Fits	Fits	127–226
glm-edge-1.5b-chat2.1 GB at Q4	Fits	Fits	Fits	127–226
LFM2-VL-1.6B2.1 GB at Q4	Fits	Fits	Fits	128–227
Falcon-H1-1.5B-Deep-Instruct2.1 GB at Q4	Fits	Fits	Fits	130–232
Falcon-H1-1.5B-Instruct2.1 GB at Q4	Fits	Fits	Fits	130–232
MiniCPM-V-4.7-1B2.1 GB at Q4	Fits	Fits	Fits	130–232
Qwen2.5-1.5B-Instruct2.1 GB at Q4	Fits	Fits	Fits	131–233
Qwen2.5-1.5B2.1 GB at Q4	Fits	Fits	Fits	131–233
Qwen2.5-Coder-1.5B-Instruct2.1 GB at Q4	Fits	Fits	Fits	131–233
Qwen2.5-Coder-1.5B2.1 GB at Q4	Fits	Fits	Fits	131–233
Qwen2.5-Math-1.5B2.1 GB at Q4	Fits	Fits	Fits	131–233
Qwen2-1.5B-Instruct2.1 GB at Q4	Fits	Fits	Fits	131–233
Qwen2.5-Math-1.5B-Instruct2.1 GB at Q4	Fits	Fits	Fits	131–233
ReaderLM-v22.1 GB at Q4	Fits	Fits	Fits	131–233
reader-lm-1.5b2.1 GB at Q4	Fits	Fits	Fits	131–233
Hymba-1.5B-Instruct2.1 GB at Q4	Fits	Fits	Fits	133–237
OLMo-2-0425-1B2.0 GB at Q4	Fits	Fits	Fits	136–243
OLMo-2-0425-1B-Instruct2.0 GB at Q4	Fits	Fits	Fits	136–243
Yi-Coder-1.5B-Chat2.0 GB at Q4	Fits	Fits	Fits	137–244
Bolmo-1B2.0 GB at Q4	Fits	Fits	Fits	138–245
granite-4.0-h-1b2.0 GB at Q4	Fits	Fits	Fits	139–246
granite-3.1-1b-a400m-instruct1.9 GB at Q4 · MoE	Fits	Fits	Fits	up to 901
granite-3.0-1b-a400m-instruct1.9 GB at Q4 · MoE	Fits	Fits	Fits	up to 901
MiniCPM-V-4.61.9 GB at Q4	Fits	Fits	Fits	156–277
MiniCPM-V-4.6-Thinking1.9 GB at Q4	Fits	Fits	Fits	156–277
kanana-2-1.3b-instruct1.9 GB at Q4	Fits	Fits	Fits	157–279
OLMo-1B-0724-hf1.9 GB at Q4	Fits	Fits	Fits	158–281
EXAONE-4.0-1.2B1.9 GB at Q4	Fits	Fits	Fits	158–282
Llama-3.2-1B-Instruct1.9 GB at Q4	Fits	Fits	Fits	164–292
Llama-3.2-1B1.9 GB at Q4	Fits	Fits	Fits	164–292
Llama-3.2-1B1.9 GB at Q4	Fits	Fits	Fits	164–292
Zamba2-1.2B-instruct1.8 GB at Q4	Fits	Fits	Fits	167–296
HRM-Text-1B1.8 GB at Q4	Fits	Fits	Fits	171–305
AMD-OLMo-1B1.8 GB at Q4	Fits	Fits	Fits	172–306
LFM2.5-1.2B-Instruct1.8 GB at Q4	Fits	Fits	Fits	173–308
LFM2-1.2B1.8 GB at Q4	Fits	Fits	Fits	173–308
LFM2.5-1.2B-Thinking1.8 GB at Q4	Fits	Fits	Fits	173–308
LFM2.5-1.2B-JP-2026061.8 GB at Q4	Fits	Fits	Fits	173–308
MiniCPM5-1B1.8 GB at Q4	Fits	Fits	Fits	188–333
MiniCPM5-1B-SFT1.8 GB at Q4	Fits	Fits	Fits	188–333
gemma-3-1b-it1.7 GB at Q4	Fits	Fits	Fits	203–360
gemma-3-1b-pt1.7 GB at Q4	Fits	Fits	Fits	203–360
Qwen3.5-0.8B1.6 GB at Q4	Fits	Fits	Fits	232–412
Falcon-E-3B-Instruct1.6 GB at Q4	Fits	Fits	Fits	234–417
bitnet-b1.58-2B-4T1.6 GB at Q4	Fits	Fits	Fits	238–424
FastVLM-0.5B1.5 GB at Q4	Fits	Fits	Fits	267–475
Qwen3-0.6B1.5 GB at Q4	Fits	Fits	Fits	270–479
LFM2-700M1.5 GB at Q4	Fits	Fits	Fits	273–485
Falcon-H1-Tiny-R-0.6B1.4 GB at Q4	Fits	Fits	Fits	326–579
Qwen3-0.6B-Base1.4 GB at Q4	Fits	Fits	Fits	340–604
d1-omni-600M1.4 GB at Q4	Fits	Fits	Fits	345–614
HyperCLOVAX-SEED-Text-Instruct-0.5B1.4 GB at Q4	Fits	Fits	Fits	358–636
Falcon-H1-0.5B-Instruct1.4 GB at Q4	Fits	Fits	Fits	389–691
Qwen2.5-0.5B-Instruct1.3 GB at Q4	Fits	Fits	Fits	410–729
Qwen2.5-0.5B1.3 GB at Q4	Fits	Fits	Fits	410–729
Qwen2-0.5B-Instruct1.3 GB at Q4	Fits	Fits	Fits	410–729
Qwen2.5-Coder-0.5B-Instruct1.3 GB at Q4	Fits	Fits	Fits	410–729
LFM2-VL-450M1.3 GB at Q4	Fits	Fits	Fits	449–799
LFM2.5-VL-450M1.3 GB at Q4	Fits	Fits	Fits	452–803
Apertus-v1.1-0.5B1.3 GB at Q4	Fits	Fits	Fits	462–822
MiniCPM4-0.5B1.3 GB at Q4	Fits	Fits	Fits	467–830
circuit-sparsity1.3 GB at Q4	Fits	Fits	Fits	483–860
SmolLM2-360M1.3 GB at Q4	Fits	Fits	Fits	560–996
SmolLM-360M1.3 GB at Q4	Fits	Fits	Fits	560–996
ERNIE-4.5-0.3B-PT1.3 GB at Q4	Fits	Fits	Fits	562–999
LFM2.5-350M1.2 GB at Q4	Fits	Fits	Fits	572–1016
LFM2-350M1.2 GB at Q4	Fits	Fits	Fits	572–1016
granite-4.0-350m1.2 GB at Q4	Fits	Fits	Fits	575–1022
granite-4.0-h-350m1.2 GB at Q4	Fits	Fits	Fits	595–1059

Copy the parts list

DGX Spark vs. Framework Desktop for local AI
GPU: NVIDIA DGX Spark 128GB — $6,950 (NVIDIA list price (not re-verified 2026-11-05: NVIDIA marketplace blocked/empty response; last confirmed $6,950 on marketplace listing; widely reported sold out in the US))
Total: $6,950
Usable memory: 96.0 GB
Built with whatcanirunai.vercel.app/planner/ (estimates; prices are dated snapshots)

Open in planner and tweak Compare with other builds