Skip to content

NVIDIA GeForce RTX 3050

NVIDIA · 8GBGDDR6 · Can run 57 models

BuyAmazon
ManufacturerNVIDIA
VRAM8 GB
Memory TypeGDDR6
ArchitectureAmpere
CUDA Cores2,560
Tensor Cores80
Bandwidth224 GB/s
TDP130W
MSRP$249
ReleasedJan 27, 2022

AI Notes

The RTX 3050 is an older budget GPU with 8GB of GDDR6 VRAM. It can run 7B-parameter models at Q4 quantization, but the Ampere-generation tensor cores and limited bandwidth make inference noticeably slower than newer alternatives. Suitable for experimenting with small models if you already own one.

Compatible Models

ModelParametersBest QuantVRAM UsedFitEst. Speed
Qwen 3 0.6B600MQ4_K_M2.5 GBRuns~90 tok/s
Qwen 3.5 0.8B800MQ4_K_M1.5 GBRuns~149 tok/s
Gemma 3 1B1BQ8_02 GBRuns~112 tok/s
Llama 3.2 1B1BQ8_03 GBRuns~75 tok/s
DeepSeek R1 1.5B1.5BQ8_03 GBRuns~75 tok/s
SmolLM2 1.7B1.7BQ8_02.7 GBRuns~83 tok/s
Gemma 2 2B2BQ8_04 GBRuns~56 tok/s
Gemma 3n E2B2BQ4_K_M3.3 GBRuns~68 tok/s
Gemma 4 E2B2BQ4_K_M4 GBRuns~56 tok/s
Qwen 3.5 2B2BQ4_K_M3 GBRuns~75 tok/s
Llama 3.2 3B3BQ8_05 GBRuns~45 tok/s
StarCoder2 3B3BQ4_K_M3.5 GBRuns~64 tok/s
Phi-3 Mini 3.8B3.8BQ8_05.8 GBRuns~39 tok/s
Phi-4 Mini 3.8B3.8BQ4_K_M4.5 GBRuns~50 tok/s
Gemma 3 4B4BQ4_K_M5 GBRuns~45 tok/s
Gemma 3n E4B4BQ4_K_M4.5 GBRuns~50 tok/s
Gemma 4 E4B4BQ4_K_M6 GBRuns~37 tok/s
Qwen 3 4B4BQ4_K_M4.5 GBRuns~50 tok/s
Qwen 3.5 4B4BQ4_K_M4.5 GBRuns~50 tok/s
Yi 1.5 6B6BQ4_K_M5 GBRuns~45 tok/s
Falcon 3 7B7BQ4_K_M6.8 GBRuns~33 tok/s
InternLM 2.5 7B7BQ4_K_M5.5 GBRuns~41 tok/s
StarCoder2 7B7BQ4_K_M5.5 GBRuns~41 tok/s
Aya Expanse 8B8BQ4_K_M6.5 GBRuns~34 tok/s
Dolphin 3 8B8BQ4_K_M6 GBRuns~37 tok/s
Nous Hermes 2 8B8BQ4_K_M6 GBRuns~37 tok/s
Yi 1.5 9B9BQ4_K_M6.5 GBRuns~34 tok/s
Codestral Mamba 7B7BQ4_K_M6.9 GBRuns (tight)~32 tok/s
OpenChat 3.5 7B7BQ4_K_M6.9 GBRuns (tight)~32 tok/s
Qwen 2.5 VL 7B7BQ4_K_M7 GBRuns (tight)~32 tok/s
WizardLM 2 7B7BQ4_K_M6.9 GBRuns (tight)~32 tok/s
Cogito 8B8BQ4_K_M7.5 GBRuns (tight)~30 tok/s
DeepSeek R1 8B8BQ4_K_M7.5 GBRuns (tight)~30 tok/s
Nemotron 3 Nano 8B8BQ4_K_M7.5 GBRuns (tight)~30 tok/s
Qwen 3 8B8BQ4_K_M7.5 GBRuns (tight)~30 tok/s
Qwen 3.5 9B9BQ4_K_M7.5 GBRuns (tight)~30 tok/s
DeepSeek R1 7B7BQ8_09 GBCPU Offload~8 tok/s
Mistral 7B7BQ8_09 GBCPU Offload~8 tok/s
Qwen 2.5 7B7BQ8_09 GBCPU Offload~8 tok/s
Qwen 2.5 Coder 7B7BQ8_09 GBCPU Offload~8 tok/s
Granite 3.3 8B8BQ8_010 GBCPU Offload~7 tok/s
Llama 3.1 8B8BQ8_010 GBCPU Offload~7 tok/s
Gemma 2 9B9BQ8_011 GBCPU Offload~6 tok/s
Yi Coder 9B9BQ4_K_M8 GBCPU Offload~8 tok/s
Falcon 3 10B10BQ4_K_M8.5 GBCPU Offload~8 tok/s
Llama 3.2 Vision 11B11BQ4_K_M8.5 GBCPU Offload~8 tok/s
Gemma 3 12B12BQ4_K_M10.5 GBCPU Offload~6 tok/s
Mistral Nemo 12B12BQ4_K_M9.5 GBCPU Offload~7 tok/s
DeepSeek R1 14B14BQ4_K_M9.9 GBCPU Offload~7 tok/s
Phi-4 14B14BQ4_K_M9.9 GBCPU Offload~7 tok/s
Phi-4 Reasoning 14B14BQ4_K_M11 GBCPU Offload~6 tok/s
Qwen 2.5 14B14BQ4_K_M9.9 GBCPU Offload~7 tok/s
Qwen 2.5 Coder 14B14BQ4_K_M12 GBCPU Offload~6 tok/s
Qwen 3 14B14BQ4_K_M12 GBCPU Offload~6 tok/s
StarCoder2 15B15BQ4_K_M10.5 GBCPU Offload~6 tok/s
InternLM 2.5 20B20BQ4_K_M12 GBCPU Offload~6 tok/s
Qwen 3.5 35B A3B35BQ4_K_M12 GBCPU Offload~6 tok/s
57 model(s) are too large for this hardware.