Skip to content

What can the 2× RTX 5090 run?

What fits on the 2× RTX 5090
Text models: 130Visual models: 0Strongest first
Memory
64 GB
Bandwidth
1,792 GB/s each
Maker
NVIDIA

Open the calculator on this machine

Text models that run on it

Text models that run on the 2× RTX 5090, best first
ModelBest fitContextSpeedStrongest at
Qwen3.8 27B8-bit256K41 tok/sChat, Coding, Agents, Docs
Muse Glimmer 30B8-bit128K39 tok/sDocs
Mistral Medium 3.5 128B2-bit40K25 tok/sChat, Coding, Agents, Docs
Qwen3.5 122B A10B3-bit80K145 tok/sChat, Coding, Agents, Docs
Nemotron 3 Super 120B A12B3-bit256K99 tok/sChat, Coding, Agents, Docs
Nemotron 3.5 Bolt 74B A9B6-bit248K89 tok/sChat, Coding, Agents, Docs
Qwen3.6 35B A3B8-bit256K200 tok/sChat, Coding, Agents, Docs
Gemma 4 31B8-bit256K35 tok/sChat, Coding, Agents, Docs
Gemma 4 26B A4B8-bit256K199 tok/sChat, Coding, Agents, Docs
GLM-4.7 Flash8-bit198K187 tok/sChat, Coding, Agents
Mistral Small 4 119B3-bit224K213 tok/sChat, Docs
gpt-oss 120B3-bit128K241 tok/sChat, Coding
Qwen3 Coder Next 80B5-bit256K215 tok/sCoding, Agents
Nemotron 3.5 Lightning 30B A3B8-bit256K175 tok/sAgents, Docs
Devstral Small 2 24BFP8224K47 tok/sCoding
LFM2.5 24B A2B8-bit128K244 tok/sAgents
Mistral Small 3.2 24B8-bit128K48 tok/sChat, Coding, Agents
Ministral 3 14B8-bit256K80 tok/sCoding, Agents, Docs
Llama 3.3 70B6-bit16K21 tok/sChat, Coding, Agents, Docs
Olmo 3.1 32B8-bit64K35 tok/sChat, Coding, Agents, Docs
Qwen3 14B8-bit40K77 tok/sNot rated
Qwen3 32B8-bit40K34 tok/sNot rated
Qwen3.6 27B8-bit256K41 tok/sText and images
Qwen2.5 Coder 14B Instruct8-bit32K75 tok/sCoding
Qwen3 Coder 30B A3B Instruct8-bit256K184 tok/sCoding
Qwen3.5 27B8-bit256K41 tok/sText and images
Qwen3.5 35B A3B8-bit256K200 tok/sText and images
NVIDIA Nemotron 3 Nano 30B A3B BF168-bit256K175 tok/sNot rated
GLM 4.5 Air3-bit48K116 tok/sNot rated
Qwen3 30B A3B8-bit40K184 tok/sNot rated
Qwen3 30B A3B Instruct 25078-bit256K184 tok/sNot rated
Laguna XS.28-bit256K202 tok/sNot rated
OTel 2.0 LLM 31B IT8-bit256K35 tok/sNot rated
Qwen2.5 32B Instruct8-bit32K34 tok/sNot rated
MiniCPM V 4.7 35B A3B8-bit256K200 tok/sText and images
Qwen2.5 14B Instruct8-bit32K75 tok/sNot rated
Laguna XS 2.18-bit256K202 tok/sNot rated
Diffusiongemma 26B A4B IT8-bit256K193 tok/sText and images
Qwen2.5 VL 32B Instruct8-bit104K34 tok/sText and images
Qwen2.5 Coder 32B Instruct8-bit32K34 tok/sCoding
Laguna S 2.13-bit120K162 tok/sNot rated
Qwen3 VL 30B A3B Instruct8-bit256K184 tok/sText and images
Gemma 3 12B IT8-bit128K92 tok/sText and images
Phi 3.5 Mini Instruct8-bit128K243 tok/sNot rated
DeepSeek R1 Distill Qwen 32B8-bit104K34 tok/sNot rated
Gemma 3 27B IT8-bit128K42 tok/sText and images
Ornith 1.5 35B A3B8-bit256K200 tok/sNot rated
Ornith 1.5 122B3-bit80K125 tok/sNot rated
GPT Neox 20B8-bit2K54 tok/sNot rated
Phi 48-bit16K76 tok/sNot rated
Phi 3 Mini 128K Instruct8-bit128K243 tok/sNot rated
Qwen3 VL 32B Instruct8-bit104K34 tok/sText and images
Pixtral 12B8-bit288K88 tok/sText and images
DeepSeek R1 Distill Qwen 14B8-bit128K75 tok/sNot rated
Qwen2.5 14B Instruct 1M8-bit232K75 tok/sNot rated
Qwen2.5 72B Instruct5-bit32K23 tok/sNot rated
Qwen3 Next 80B A3B Instruct5-bit224K215 tok/sNot rated
DeepSeek V2 Lite8-bit160K258 tok/sNot rated
Llama 4 Scout 17B 16E Instruct3-bit48K101 tok/sText and images
Llama Guard 4 12B8-bit248K93 tok/sText and images

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 130 are shown.

Only just too big

These need up to a quarter more memory than the 2× RTX 5090 has. Spilling into system RAM or a lower precision can get them running, slowly.

Questions

What is the best LLM for the 2× RTX 5090?

Of the open models we track, the strongest that run well on the 2× RTX 5090 are Qwen3.8 27B, Muse Glimmer 30B, Mistral Medium 3.5 128B. 130 text models fit it in all.

Calculated from published specs, not benchmarked. All machines