NVIDIA · consumer

GeForce RTX 3050

GeForce RTX 3050 has 8 GB of VRAM at 224 GB/s — about 7.44 GiB usable after driver and compositor overhead. 1407 of 2118 indexed models fit at 16K context with q4_0 KV.

Spec sheet· bandwidth, theoreticalFrom the file· fit from summed bytesPredicted· speed
Memory
8 GB
GDDR6
Bandwidth
224 GB/s
128-bit bus
Tensor FP16
36 TF
dense
TDP
130 W
$249 MSRP
Context4K8K16K32K64K128K
KV cachef16q8_0q4_0quantizing the KV cache is a ~2× lever on the dominant term at long context
text 1210vision language 102image 2video 8embedding 26audio tts 21audio asr 38

What fits at 16K context

largest quantization that fits, per model · 1407 of 2118 indexed
ModelBest quantParamsWeights●KV●Total in memory◐Headroom◐tok/s◐
SuperGemma-4-12b-abliteratedI1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
gemma-4-12B-agentic-fable5-composer2.5-v2-3.5x-tau2-uncensored-hereticI1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
gemma-4-12B-coder-fable5-composer2.5-v1-uncensored-hereticI1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
gemma-4-12B-it-uncensored-hereticI1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
Grug-12BI1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
Aura-Medium-v1-BF16I1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
gemma-4-12B-it-Esper4I1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
gemma-4-12B-it-GuardpointI1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
Gemma-4-12B-it-AEON-Abliterated-K4-BF16I1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
gemma-4-12B-it-Tachibana-AgentI1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
gemma-4-12b-marvin-gutenberg-rp-v2I1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
gemma-4-12b-crownelius-writerI1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
Huihui-gemma-4-12B-coder-fable5-composer2.5-v1-abliteratedI1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
gemma-4-12b-asterion-agenticI1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
Huihui-gemma-4-12B-agentic-fable5-abliteratedI1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
g4-12b-it-trismegistusI1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
gemma4-12b-it-asimovI1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
FabGemmaI1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
Huihui-gemma-4-12B-it-qat-q4_0-unquantized-abliteratedI1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
gemma-4-12B-it-abliterated-uncensoredI1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
Gemma-4-12b-it-AbliteratedI1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
gemma-4-12B-Queen-it-qat-q4_0-unquantizedI1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
gemma-4-12B-it-heretic_decensoredI1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
Iris-12B-gemma-4-it-qatI1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
gemma-4-12B-coder-fable5-composer2.5-v1I1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
G4-Starry-Ocean-12BI1-IQ4_XS11.9B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
gemma-4-12B-it-QAT-SOMPOA-heresyI1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
gemma-4-12B-it-uncensored-opus4.7-cotI1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
Gemma4-12B-IT-AbliteratedI1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
gemma-4-12b-it-uncensoredI1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
Huihui-gemma-4-12B-it-abliteratedI1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
gemma-4-12B-it-hereticI1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
Tema_Q-X5-12B-ThinkingI1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
gemma-4-12B-coder-fable5-composer2.5-v1-bf16I1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
swarm-sovereign-12bI1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
Gemma-4-12B-OBLITERATEDI1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
Gemma4-12B-UncensoredI1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
Serenity-12BI1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
Dark-PaneI1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
Reelva-12BI1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
G4-Starry-Ocean-12B-hereticI1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
Iris-12B-v1.3.2I1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
Semancer-12BI1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
Iris-12B-v1.2I1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
gemma-4-12b-heretic-abliteratedI1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
gemma-4-12BIQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
gemma-4-12b-marvin-gutenbergI1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
gemma-4-12b-marvin-v2I1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
gemma-4-12b-marvin-v1I1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
STARK-WEB-12B-v1.7I1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
STARK-WEB-12BI1-IQ4_XS12.0B6.18 GiB0.41 GiB7.44 GiB0.00 GiB25±12.9%
OmniAtlas-Qwen3-30B-A3BI1-IQ1_M31.7B6.59 GiB0.00 GiB7.44 GiB0.00 GiB25±12.9%
Qwen3-Omni-30B-A3B-CaptionerI1-IQ1_M31.7B6.59 GiB0.00 GiB7.44 GiB0.00 GiB25±12.9%
GLM-4.7-Flash-DerestrictedMoEI1-IQ1_M31.2B6.39 GiB0.23 GiB7.43 GiB0.01 GiB88±37%
Huihui-GLM-4.7-Flash-abliteratedMoEI1-IQ1_M31.2B6.39 GiB0.23 GiB7.43 GiB0.01 GiB88±37%
Mistral-NeMo-Minitron-8B-InstructQ5_K_L8.4B5.90 GiB0.70 GiB7.43 GiB0.01 GiB24±12.9%
DeepSeek-Coder-V2-Lite-BaseMoEI1-IQ3_XXS15.7B6.49 GiB0.13 GiB7.43 GiB0.01 GiB81±37%
DeepSeek-Coder-V2-Lite-InstructMoEIQ3_XXS15.7B6.49 GiB0.13 GiB7.43 GiB0.01 GiB81±37%
DeepSeek-V2-Lite-ChatMoEIQ3_XXS15.7B6.49 GiB0.13 GiB7.43 GiB0.01 GiB81±37%
Cydonia-v1.3-Magnum-v4-22BI1-IQ2_XXS22.2B5.58 GiB0.98 GiB7.43 GiB0.01 GiB25±12.9%
●From the file◐Predictedwhat these mean

Speed is modeled, not measured: decode is memory-bandwidth bound, so tokens per second is bytes read per token against achievable bandwidth. Mixture-of-experts models carry a wider band because only the routed experts are read each step, and few have been measured publicly.

Questions people ask

What AI models can a GeForce RTX 3050 run?
1407 of 2118 indexed open-weight models fit a GeForce RTX 3050 at 16,384 context with q4_0 KV cache, the largest being SuperGemma-4-12b-abliterated at I1-IQ4_XS. That covers text, vision-language, image, video and speech models.
How much usable memory does a GeForce RTX 3050 actually have?
Its nameplate is 8 GB, but about 7.44 GiB is available to a model once driver and compositor overhead is accounted for.
Is a GeForce RTX 3050 fast for local AI?
Its memory bandwidth is 224 GB/s, and that figure — not teraflops — is what governs token generation speed. Capacity decides what you can run; bandwidth decides how fast it runs.