Local AI Advisor

What LLMs can Desktop PC · RTX PRO 6000 Blackwell · 128 GB RAM run?

96 GB VRAM + 128 GB RAM, 1792 GB/s memory bandwidth, about $17,284. Every open model in our catalog, rated by the engine for chat and for agentic coding.

Best for chat

Qwen3.6 35B-A3B

Excellent · Q6_K · 260–360 tok/s

Best for agentic coding

Qwen3.6 35B-A3B

Excellent · Q6_K · 230–310 tok/s

Best open-source (OSI) pick

Qwen3.6 35B-A3B

Excellent · Q6_K · 260–360 tok/s

Runs well (40)

Fast enough and with memory to spare for everyday chat.

ModelChatAgentic codingQuantChat speed
Qwen3.6 35B-A3BAlibaba QwenOpen sourceExcellentExcellentQ6_K260–360 tok/s
Qwen3-Next 80B-A3B InstructAlibaba QwenOpen sourceExcellentComfortableQ6_K260–360 tok/s
gpt-oss-120bOpenAIOpen sourceExcellentExcellentMXFP4≈180 tok/s
Gemma 4 26B-A4BGoogleOpen sourceExcellentComfortableQ6_K210–280 tok/s
Qwen3.5 35B-A3BAlibaba QwenOpen sourceExcellentComfortableQ6_K260–360 tok/s
LFM2.5 24B-A2BLiquid AIExcellentComfortableQ4_K_M530–720 tok/s
gpt-oss-20bOpenAIOpen sourceExcellentComfortableMXFP4310–420 tok/s
Qwen3.5 122B-A10BAlibaba QwenOpen sourceExcellentExcellentQ5_K_M95–130 tok/s
GLM-4.7-Flash (30B-A3B)Z.aiOpen sourceExcellentComfortableQ6_K260–350 tok/s
Mistral Small 4 (119B-A6.5B)Mistral AIOpen sourceExcellentComfortableQ5_K_M140–190 tok/s
Qwen3 30B-A3B Instruct 2507Alibaba QwenOpen sourceExcellentComfortableQ6_K230–310 tok/s
Kolibri-1 (78B-A3.5B)Aleph AlphaOpen sourceExcellentAcceptableFP8190–260 tok/s
Qwen3.8 27BAlibaba QwenOpen sourceExcellentExcellentQ6_K40–50 tok/s
GLM-4.5-Air (106B-A12B)Z.aiOpen sourceExcellentComfortableQ5_K_M75–100 tok/s
Qwen3.6 27BAlibaba QwenOpen sourceExcellentExcellentQ6_K40–50 tok/s
Qwen3-Coder 30B-A3B InstructAlibaba QwenOpen sourceExcellentComfortableQ6_K230–310 tok/s
MiniCPM5 2BOpenBMBOpen sourceExcellentAcceptableQ8_0320–440 tok/s
Gemma 4 31BGoogleOpen sourceExcellentComfortableQ6_K35–45 tok/s
Qwen3.5 27BAlibaba QwenOpen sourceExcellentExcellentQ6_K40–50 tok/s
Qwen3.5 9BAlibaba QwenOpen sourceExcellentAcceptableQ6_K110–150 tok/s
Qwen3.5 4BAlibaba QwenOpen sourceExcellentAcceptableQ8_0200–270 tok/s
Gemma 4 12BGoogleOpen sourceExcellentAcceptableQ6_K90–120 tok/s
Nemotron 3.5 Bolt 74B-A9BNVIDIAExcellentAcceptable4-bit130–180 tok/s
GLM-4.6V-Flash (10B)Z.aiOpen sourceExcellentAcceptableQ6_K100–140 tok/s
Qwen3 8BAlibaba QwenOpen sourceExcellentAcceptableQ6_K120–170 tok/s
Qwen3 14BAlibaba QwenOpen sourceExcellentAcceptableQ6_K70–95 tok/s
DeepSeek-R1-0528-Qwen3-8BDeepSeekOpen sourceExcellentAcceptableQ6_K120–170 tok/s
Muse Glimmer (30B)MetaOpen sourceExcellentAcceptableQ4_K_M55–70 tok/s
Gemma 4 E4BGoogleOpen sourceExcellentBorderlineQ6_K130–180 tok/s
Llama 4 Scout (109B-A17B)MetaOpen weights (restricted)ExcellentAcceptableQ5_K_M55–70 tok/s
Mistral Small 3.2 (24B)Mistral AIOpen sourceExcellentAcceptableQ6_K45–60 tok/s
Phi-4 (14B)MicrosoftOpen sourceExcellentAcceptableQ6_K70–95 tok/s
Gemma 3 27BGoogleOpen weights (restricted)ExcellentAcceptableQ6_K40–55 tok/s
Llama 3.2 3B InstructMetaOpen weights (restricted)ExcellentTechnically RunsQ6_K300–410 tok/s
Llama 3.1 8B InstructMetaOpen weights (restricted)ExcellentBorderlineQ6_K130–170 tok/s
Qwen3 32BAlibaba QwenOpen sourceExcellentComfortableQ6_K30–45 tok/s
Devstral Small 2 (24B)Mistral AIOpen sourceExcellentExcellentQ6_K45–60 tok/s
Qwen2.5-Coder 7B InstructAlibaba QwenOpen sourceExcellentAcceptableQ6_K140–180 tok/s
Qwen2.5-Coder 32B InstructAlibaba QwenOpen sourceExcellentAcceptableQ6_K30–45 tok/s
Llama 3.3 70B InstructMetaOpen weights (restricted)ExcellentAcceptableQ5_K_M17–25 tok/s

Loads, but not recommended (2)

Technically runs. Expect a frustrating experience.

ModelChatAgentic codingQuantChat speed
DeepSeek-V4-Flash (284B-A13B)DeepSeekOpen sourceBorderlineBorderlineQ4_K_M7–10 tok/s
GLM-5.3-Flash (320B-A18B)Z.aiOpen sourceBorderlineBorderlineQ4_K_M4.3–6 tok/s

Doesn't fit or isn't supported (6)

Not enough usable memory, or the runtime can't run it here.

ModelChatAgentic codingQuantChat speed
DeepSeek-V4.1-Flash (748B-A16B)DeepSeekOpen sourceDoes Not FitDoes Not Fit——
Muse Spark Open 816B-A42BMetaDoes Not FitDoes Not Fit——
GLM-5.3 (753B-A40B)Z.aiDoes Not FitDoes Not Fit——
Nemotron 3.5 Super VL 120B-A12BNVIDIAUnsupportedUnsupported——
Reflection Beam (501B-A23B)Reflection AIOpen sourceUnsupportedUnsupported——
Step 5 Preview 600B-A27BStepfunUnsupportedUnsupported——

Want a specific tool, context size or workload?

The full check models your exact setup, including the apps you keep open.

Check this machine in detail