BuildsGPUsModelsGuidesMethodologyNewsletter

What local AI models can the NVIDIA GeForce RTX 3090 Ti run?

24 GB GDDR6X, 1010 GB/s, 450 W TDP. Specs: TechPowerUp (accessed 2026-10-15). Price: check price

Buy on Amazon As an Amazon Associate I earn from qualifying purchases.

Model (Q4)FileFits?Speed estimate
Qwen3-0.6B0.48 GB✅~1599 tok/s
Llama-3.2-3B-Instruct2.02 GB✅~380 tok/s
gemma-3-4b-it2.49 GB✅~308 tok/s
Qwen3-4B2.5 GB✅~307 tok/s
Mistral-7B-Instruct-v0.34.37 GB✅~176 tok/s
Qwen2.5-Coder-7B-Instruct4.68 GB✅~164 tok/s
Llama-3.1-8B-Instruct4.92 GB✅~156 tok/s
Qwen3-8B5.03 GB✅~153 tok/s
NVIDIA-Nemotron-Nano-9B-v26.53 GB✅~118 tok/s
gemma-3-12b-it7.3 GB✅~105 tok/s
Mistral-Nemo-Instruct-24077.48 GB✅~103 tok/s
Qwen2.5-Coder-14B-Instruct8.99 GB✅~85 tok/s
DeepSeek-R1-Distill-Qwen-14B8.99 GB✅~85 tok/s
Qwen3-14B9 GB✅~85 tok/s
phi-49.05 GB✅~85 tok/s
gpt-oss-20b11.67 GB✅~382 tok/s
Mistral-Small-3.2-24B-Instruct-250614.33 GB✅~54 tok/s
Devstral-Small-250514.33 GB✅~54 tok/s
gemma-3-27b-it16.55 GB✅~46 tok/s
Qwen3-Coder-30B-A3B-Instruct18.56 GB✅~383 tok/s
Qwen3-30B-A3B18.63 GB✅~381 tok/s
Qwen3-32B19.76 GB✅~39 tok/s
Qwen2.5-Coder-32B-Instruct19.85 GB✅~39 tok/s
DeepSeek-R1-Distill-Qwen-32B19.85 GB✅~39 tok/s
Llama-3.3-70B-Instruct42.52 GB❌—
DeepSeek-R1-Distill-Llama-70B42.52 GB❌—
Qwen2.5-72B-Instruct47.42 GB❌—
gpt-oss-120b62.84 GB❌—
GLM-4.5-Air73.5 GB❌—
Qwen3-235B-A22B142.15 GB❌—

Speed calculator estimate

Estimate = 0.76 × bandwidth ÷ GB read per token (Q4 file size, scaled by active/total parameters for MoE). Fit = file size + 1.5 GB headroom ≤ total VRAM. Multi-GPU: VRAM adds, speed uses one card's bandwidth (layer split). Short context only. How this works