NVIDIA · consumer

GeForce RTX 3050 OEM

GeForce RTX 3050 OEM has 8 GB of VRAM at 224 GB/s — about 7.44 GiB usable after driver and compositor overhead. 1345 of 2118 indexed models fit at 32K 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
32 TF
dense
TDP
130 W
Context4K8K16K32K64K128K
KV cachef16q8_0q4_0quantizing the KV cache is a ~2× lever on the dominant term at long context
text 1152vision language 98image 2video 8embedding 26audio tts 21audio asr 38

What fits at 32K context

largest quantization that fits, per model · 1345 of 2118 indexed
ModelBest quantParamsWeights●KV●Total in memory◐Headroom◐tok/s◐
L3.2-Rogue-Creative-Instruct-Uncensored-Abliterated-7BQ4_K_M7.5B4.27 GiB2.36 GiB7.44 GiB0.00 GiB24±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%
granite-4.1-8bQ4_18.8B5.20 GiB1.41 GiB7.44 GiB0.00 GiB24±12.9%
Qwen3-VL-8B-Instruct-HereticI1-IQ2_S8.8B5.34 GiB1.27 GiB7.43 GiB0.01 GiB24±12.9%
Forsaken-Void-12BI1-IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
Silver-Siren-ST-12BI1-IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
Tess-3-Mistral-Nemo-12BI1-IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
KrakenSakura-Maelstrom-12B-v1IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
MN-12B-Runeweaver-RP-RUI1-IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
Dans-PersonalityEngine-V1.3.0-12bI1-IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
Impish_Bloodmoon_12BI1-IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
Wayfarer-2-12BI1-IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
Wayfarer-12BI1-IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
Muse-12BI1-IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
mini-magnum-12b-v1.1IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
MN-Violet-Lotus-12BI1-IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
Rocinante-X-12B-v1-Heretic-UncensoredI1-IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
Mistral-Heretica-12BI1-IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
arcee-fusion-lumaid-12BI1-IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
Mistral-NeMo-12B-AbliteratedI1-IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
Captain-Eris_Violet-V0.420-12BI1-IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
Rocinante-X-12B-v1-absolute-heresyI1-IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
Rocinante-X-12B-v1I1-IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
Mistral-Nemo-2407-12B-Thinking-Claude-Gemini-GPT5.2-Uncensored-HERETICI1-IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
Dans-SakuraKaze-V1.0.0-12bI1-IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
Mistral-Nemo-Inst-2407-12B-Thinking-Uncensored-HERETIC-HI-Claude-OpusI1-IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
Mistral-Nemo-Instruct-2407-12B-Thinking-M-Claude-Opus-High-ReasoningI1-IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
MN-12B-Mag-Mell-R1IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
Mordant-12B-ThinkI1-IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
Riverfish-Rocinante-12B-SFT-DPOI1-IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
MN-Violet-Lotus-12B-HereticI1-IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
Himeyuri-Magnum-12B-HereticMergeI1-IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
Kinggaroo-12b-v1I1-IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
Magnum-Picaro-0.7-v2-12bI1-IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
magnum-v2.5-12b-ktoI1-IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
Mistral-Nemo-Prism-12BI1-IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
Nera_Noctis-12BI1-IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
Chronos-Gold-12B-1.0I1-IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
patricide-12B-Unslop-MellI1-IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
MN-12B-Celeste-V1.9I1-IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
magnum-v2-12bI1-IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
Peaceful-Days-12BI1-IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
NemoMix-Unleashed-12BI1-IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
Mistral-Nemo-Base-2407I1-IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
Blissful-Days-12BI1-IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
Rocinante-12B-v1.1I1-IQ3_S12.2B5.18 GiB1.41 GiB7.43 GiB0.01 GiB25±12.9%
NVIDIA-Nemotron-Nano-9B-v2IQ2_S8.9B4.62 GiB1.97 GiB7.43 GiB0.01 GiB25±12.9%
granite-8b-code-instruct-4kI1-Q5_K_M8.1B5.33 GiB1.27 GiB7.43 GiB0.01 GiB25±12.9%
granite-8b-code-base-4kI1-Q5_K_M8.1B5.33 GiB1.27 GiB7.43 GiB0.01 GiB25±12.9%
NVIDIA-Nemotron-Nano-12B-v2Q2_K12.3B4.38 GiB2.18 GiB7.43 GiB0.01 GiB25±12.9%
gemma-7bI1-IQ2_XS8.5B2.62 GiB3.94 GiB7.43 GiB0.01 GiB25±12.9%
Ministral-8B-Instruct-2410Q5_K_M8.0B5.33 GiB1.27 GiB7.43 GiB0.01 GiB24±12.9%
GLM-4.6V-FlashQ5_K_S10.3B6.24 GiB0.35 GiB7.43 GiB0.01 GiB25±12.9%
GLM-Z1-9B-0414Q5_K_S9.4B6.24 GiB0.35 GiB7.43 GiB0.01 GiB25±12.9%
glm4.1v-9b-base-sftI1-Q5_K_S10.3B6.24 GiB0.35 GiB7.43 GiB0.01 GiB25±12.9%
GLM-4-9B-0414Q5_K_S9.4B6.24 GiB0.35 GiB7.43 GiB0.01 GiB25±12.9%
GLM-4.1V-9B-ThinkingQ5_K_S10.3B6.24 GiB0.35 GiB7.43 GiB0.01 GiB25±12.9%
next-8bI1-Q5_K_S8.2B5.33 GiB1.27 GiB7.43 GiB0.01 GiB25±12.9%
Supertron2-Reranker-8BI1-Q5_K_S8.8B5.33 GiB1.27 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 OEM run?
1345 of 2118 indexed open-weight models fit a GeForce RTX 3050 OEM at 32,768 context with q4_0 KV cache, the largest being L3.2-Rogue-Creative-Instruct-Uncensored-Abliterated-7B at Q4_K_M. That covers text, vision-language, image, video and speech models.
How much usable memory does a GeForce RTX 3050 OEM 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 OEM 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.