Skip to content

What can the 2× RTX PRO 6000 Blackwell run?

What fits on the 2× RTX PRO 6000 Blackwell
Text models: 80Visual models: 0Strongest first
Memory
192 GB
Bandwidth
1,792 GB/s each
Maker
NVIDIA

Open the calculator on this machineRTX PRO 6000 Blackwell in our inventory

Text models that run on it

Text models that run on the 2× RTX PRO 6000 Blackwell, best first
ModelBest fitContextSpeedStrongest at
DeepSeek V4.1 Flash1.6-bit1M193 tok/sChat, Coding, Agents, Docs
GLM-5.31.6-bit352K89 tok/sChat, Coding, Agents, Docs
GLM-5.3 Flash3-bit1M102 tok/sChat, Coding, Agents, Docs
Qwen3.8 Flash-Next8-bit104K124 tok/sChat, Coding, Agents
Muse Spark Open 816B A42B1.6-bit352K94 tok/sChat, Docs
MiMo V2.6 FlashFP4 + FP8896K87 tok/sCoding, Agents
MiniMax M32-bit256K95 tok/sDocs
DeepSeek V4 FlashFP4 + FP81M57 tok/sDocs
Hy34-bit44K67 tok/sChat, Coding, Agents, Docs
Nemotron 3 Ultra 550B A55B1.6-bit256K63 tok/sChat, Coding, Docs
Qwen3.5 397B A17B2-bit256K113 tok/sAgents, Docs
MiniMax M2.76-bit24K97 tok/sCoding, Agents
DeepSeek V3.21.6-bit160K103 tok/sChat, Coding, Agents, Docs
Mistral Medium 3.5 128BFP8160K9.3 tok/sChat, Coding, Agents, Docs
Qwen3.5 122B A10B8-bit256K87 tok/sChat, Coding, Agents, Docs
Nemotron 3 Super 120B A12B8-bit256K64 tok/sChat, Coding, Agents, Docs
Nemotron 3.5 Bolt 74B A9B8-bit256K77 tok/sChat, Coding, Agents, Docs
Mistral Small 4 119BFP8256K135 tok/sChat, Docs
gpt-oss 120BMXFP4128K197 tok/sChat, Coding
Qwen3 Coder Next 80B8-bit256K182 tok/sCoding, Agents
Llama 3.3 70B8-bit128K16 tok/sChat, Coding, Agents, Docs
Inkling1.6-bit4K94 tok/sText and images
DeepSeek R11.6-bit160K98 tok/sNot rated
DeepSeek V31.6-bit160K98 tok/sNot rated
GLM 51.6-bit198K86 tok/sNot rated
GLM 4.5 Air8-bit128K65 tok/sNot rated
Inkling Small5-bit4K111 tok/sText and images
MiniMax M2.56-bit24K89 tok/sNot rated
GLM 5.11.6-bit198K86 tok/sNot rated
Laguna M.16-bit28K51 tok/sNot rated
Laguna S 2.18-bit1M100 tok/sNot rated
Qwen3 VL 235B A22B Instruct5-bit128K55 tok/sText and images
GLM 5.21.6-bit352K87 tok/sNot rated
Qwen3 235B A22B5-bit40K55 tok/sNot rated
Ornith 1.5 122B8-bit256K72 tok/sNot rated
Ornith 1.5 397B2-bit256K91 tok/sNot rated
Ornith 1.0 397B2-bit256K110 tok/sNot rated
Step 3.5 Flash6-bit72K113 tok/sNot rated
DeepSeek V3 03241.6-bit160K97 tok/sNot rated
Qwen2.5 72B Instruct8-bit32K16 tok/sNot rated
Qwen3 Next 80B A3B Instruct8-bit256K182 tok/sNot rated
DeepSeek R1 05281.6-bit160K97 tok/sNot rated
GLM 4.53-bit48K52 tok/sNot rated
Llama 4 Scout 17B 16E Instruct8-bit384K55 tok/sText and images
MiniMax M26-bit24K89 tok/sNot rated
MiMo V2.54-bit240K88 tok/sNot rated
Qwen2.5 VL 72B Instruct8-bit125K16 tok/sText and images
Solar Open 100B8-bit128K76 tok/sNot rated
GLM 5.3 Flash BF163-bit1M100 tok/sText and images
Qwen3 VL 235B A22B Thinking5-bit128K55 tok/sText and images
Qwen3 235B A22B Instruct 25075-bit136K55 tok/sNot rated
NVIDIA Nemotron 3.5 Super VL 120B A12B BF168-bit1M64 tok/sText and images
Hy3 Preview4-bit36K67 tok/sNot rated
Qwen3 Coder 480B A35B Instruct2-bit52K71 tok/sCoding
GLM 4.73-bit48K52 tok/sNot rated
Kimi Linear 48B A3B Instruct8-bit4K238 tok/sNot rated
Qwen3 Next 80B A3B Thinking8-bit256K182 tok/sNot rated
GLM 5.3 BF161.6-bit352K87 tok/sNot rated
GLM 4.63-bit48K53 tok/sNot rated
DeepSeek R1 Distill Llama 70B8-bit128K16 tok/sNot rated

Best fit is the highest quality, up to 8-bit, that leaves the use case’s context. Ranked by our use-case scores, then by Hugging Face downloads. Speeds are rough, plus or minus 25%. The top 60 of 80 are shown.

Only just too big

These need up to a quarter more memory than the 2× RTX PRO 6000 Blackwell has. Spilling into system RAM or a lower precision can get them running, slowly.

Questions

What is the best LLM for the 2× RTX PRO 6000 Blackwell?

Of the open models we track, the strongest that run well on the 2× RTX PRO 6000 Blackwell are DeepSeek V4.1 Flash, GLM-5.3, GLM-5.3 Flash. 80 text models fit it in all.

Calculated from published specs, not benchmarked. All machines