NVIDIA · consumer

GeForce RTX 3080

GeForce RTX 3080 has 10 GB of VRAM at 760 GB/s — about 9.30 GiB usable after driver and compositor overhead. 1535 of 2118 indexed models fit at 8K context with f16 KV.

Spec sheet· bandwidth, theoreticalFrom the file· fit from summed bytesPredicted· speed
Memory
10 GB
GDDR6X
Bandwidth
760 GB/s
320-bit bus
Tensor FP16
119 TF
dense
TDP
320 W
$699 MSRP
Context4K8K16K32K64K128K
KV cachef16q8_0q4_0quantizing the KV cache is a ~2× lever on the dominant term at long context
text 1320vision language 115image 2video 12audio asr 39audio tts 21embedding 26

What fits at 8K context

largest quantization that fits, per model · 1535 of 2118 indexed
ModelBest quantParamsWeights●KV●Total in memory◐Headroom◐tok/s◐
Snowpiercer-15B-v4-hereticI1-Q3_K_M15.0B6.89 GiB1.56 GiB9.30 GiB0.00 GiB63±12.9%
Snowpiercer-15B-v4Q3_K_M15.0B6.89 GiB1.56 GiB9.30 GiB0.00 GiB63±12.9%
OmniAtlas-Qwen3-30B-A3BI1-IQ2_XS31.7B8.45 GiB0.00 GiB9.30 GiB0.00 GiB63±12.9%
Qwen3-Omni-30B-A3B-CaptionerI1-IQ2_XS31.7B8.45 GiB0.00 GiB9.30 GiB0.00 GiB63±12.9%
v6-Finch-7B-HFQ4_07.6B4.45 GiB4.00 GiB9.29 GiB0.01 GiB63±12.9%
rwkv-6-world-7bQ4_07.6B4.45 GiB4.00 GiB9.29 GiB0.01 GiB63±12.9%
Qwen3-Coder-REAP-25B-A3BMoEIQ2_M24.9B7.75 GiB0.75 GiB9.29 GiB0.01 GiB158±37%
Tiger-Gemma-9B-v3Q5_K_L9.2B6.40 GiB2.05 GiB9.29 GiB0.01 GiB63±12.9%
Gemma-2-9B-It-SPPO-Iter3Q5_K_L9.2B6.40 GiB2.05 GiB9.29 GiB0.01 GiB63±12.9%
gemma-2-9b-it-abliteratedQ5_K_L9.2B6.40 GiB2.05 GiB9.29 GiB0.01 GiB63±12.9%
gemma-2-9b-itQ5_K_L9.2B6.40 GiB2.05 GiB9.29 GiB0.01 GiB63±12.9%
Tiger-Gemma-9B-v1Q5_K_L9.2B6.40 GiB2.05 GiB9.29 GiB0.01 GiB63±12.9%
magnum-v4-9bQ5_K_L9.2B6.40 GiB2.05 GiB9.29 GiB0.01 GiB63±12.9%
Tini-Cybersec-8B-A1BMoEQ8_08.5B8.39 GiB0.09 GiB9.28 GiB0.02 GiB175±37%
LFM2.5-8B-A1B-KO-SFTMoEQ8_08.5B8.39 GiB0.09 GiB9.28 GiB0.02 GiB175±37%
LFM2.5-8B-A1B-SOMPOA-heresyMoEQ8_08.5B8.39 GiB0.09 GiB9.28 GiB0.02 GiB175±37%
Huihui-LFM2.5-8B-A1B-abliteratedMoEQ8_08.5B8.39 GiB0.09 GiB9.28 GiB0.02 GiB175±37%
LFM2.5-8B-A1BMoEQ8_08.5B8.39 GiB0.09 GiB9.28 GiB0.02 GiB175±37%
Supertron2.1-8B-A1BMoEQ8_08.5B8.39 GiB0.09 GiB9.28 GiB0.02 GiB175±37%
LFM2.5-8B-A1B-hereticMoEQ8_08.5B8.39 GiB0.09 GiB9.28 GiB0.02 GiB175±37%
deepseek-coder-6.7b-instructQ5_K_M6.7B4.46 GiB4.00 GiB9.28 GiB0.02 GiB63±12.9%
deepseek-coder-6.7b-baseQ5_K_M6.7B4.46 GiB4.00 GiB9.28 GiB0.02 GiB63±12.9%
deepseek-coder-6.7B-kexerI1-Q5_K_M6.7B4.46 GiB4.00 GiB9.28 GiB0.02 GiB63±12.9%
Magicoder-S-DS-6.7BI1-Q5_K_M6.7B4.46 GiB4.00 GiB9.28 GiB0.02 GiB63±12.9%
Phi-4-reasoningQ3_K_M14.7B6.86 GiB1.56 GiB9.28 GiB0.02 GiB63±12.9%
Phi-4-reasoning-plusQ3_K_M14.7B6.86 GiB1.56 GiB9.28 GiB0.02 GiB63±12.9%
phi-4Q3_K_M14.7B6.86 GiB1.56 GiB9.28 GiB0.02 GiB63±12.9%
Olmo-3-7B-InstructQ6_K_L7.3B5.77 GiB2.69 GiB9.28 GiB0.02 GiB63±12.9%
MathCoder2-CodeLlama-7BQ5_K_M6.7B4.45 GiB4.00 GiB9.28 GiB0.02 GiB63±12.9%
CodeLlama-7b-instruct-hfQ5_K_M6.7B4.45 GiB4.00 GiB9.28 GiB0.02 GiB63±12.9%
CodeLlama-7b-hfQ5_K_M6.7B4.45 GiB4.00 GiB9.28 GiB0.02 GiB63±12.9%
WizardLM-7B-UncensoredI1-Q5_K_M6.7B4.45 GiB4.00 GiB9.28 GiB0.02 GiB63±12.9%
Llama-2-7B-32K-InstructI1-Q5_K_M6.7B4.45 GiB4.00 GiB9.28 GiB0.02 GiB63±12.9%
Luna-AI-Llama2-UncensoredI1-Q5_K_M6.7B4.45 GiB4.00 GiB9.28 GiB0.02 GiB63±12.9%
Swallow-7b-NVE-instruct-hfI1-Q5_K_M6.7B4.45 GiB4.00 GiB9.28 GiB0.02 GiB63±12.9%
llava-v1.5-7bQ5_K_M6.7B4.45 GiB4.00 GiB9.28 GiB0.02 GiB63±12.9%
CodeLlama-7b-python-hfQ5_K_M6.7B4.45 GiB4.00 GiB9.28 GiB0.02 GiB63±12.9%
Wizard-Vicuna-7B-UncensoredQ5_K_M6.7B4.45 GiB4.00 GiB9.28 GiB0.02 GiB63±12.9%
llama2_7b_chat_uncensoredQ5_K_M6.7B4.45 GiB4.00 GiB9.28 GiB0.02 GiB63±12.9%
WizardLM-7B-V1.0-UncensoredQ5_K_M6.7B4.45 GiB4.00 GiB9.28 GiB0.02 GiB63±12.9%
Llama-2-7b-hfQ5_K_M6.7B4.45 GiB4.00 GiB9.28 GiB0.02 GiB63±12.9%
pygmalion-2-7bQ5_K_M6.7B4.45 GiB4.00 GiB9.28 GiB0.02 GiB63±12.9%
Nanbeige4.2-3BBF164.2B7.77 GiB0.69 GiB9.28 GiB0.02 GiB63±12.9%
OLMo-2-1124-7B-InstructQ4_K_L7.3B4.45 GiB4.00 GiB9.28 GiB0.02 GiB63±12.9%
Skywork-R1V3-38BIQ2_XXS38.4B8.41 GiB0.00 GiB9.27 GiB0.03 GiB63±12.9%
Hy-MT2-7BQ8_08.0B7.43 GiB1.00 GiB9.27 GiB0.03 GiB63±12.9%
medgemma-27b-itI1-IQ2_XXS28.8B7.16 GiB1.23 GiB9.27 GiB0.03 GiB63±12.9%
gemma-3-27b-it-abliterated-refined-visionI1-IQ2_XXS27.4B7.16 GiB1.23 GiB9.27 GiB0.03 GiB63±12.9%
Nidum-Gemma-3-27B-it-UncensoredI1-IQ2_XXS27.4B7.16 GiB1.23 GiB9.27 GiB0.03 GiB63±12.9%
AtomicGPT-gemma3-27bI1-IQ2_XXS27.4B7.16 GiB1.23 GiB9.27 GiB0.03 GiB63±12.9%
Unbound-v1.12.0-27BI1-IQ2_XXS27.4B7.16 GiB1.23 GiB9.27 GiB0.03 GiB63±12.9%
Mira-v1.12-Ties-27BI1-IQ2_XXS27.4B7.16 GiB1.23 GiB9.27 GiB0.03 GiB63±12.9%
Medgamma27BI1-IQ2_XXS27.0B7.16 GiB1.23 GiB9.27 GiB0.03 GiB63±12.9%
Hunyuan-7B-InstructQ8_07.5B7.43 GiB1.00 GiB9.27 GiB0.03 GiB63±12.9%
HunyuanImage-2.1Q3_K_S17.5B8.42 GiB0.00 GiB9.27 GiB0.03 GiB63±12.9%
Kimi-VL-A3B-Thinking-2506MoEIQ4_XS16.4B8.21 GiB0.24 GiB9.27 GiB0.03 GiB188±37%
Kimi-VL-A3B-InstructMoEIQ4_XS16.4B8.21 GiB0.24 GiB9.27 GiB0.03 GiB188±37%
dolphin-2.9.2-Phi-3-MediumKV unresolvedQ3_K_L14.0B6.84 GiB1.56 GiB9.27 GiB0.03 GiB63±12.9%
Rocinante-XL-16B-v1I1-IQ3_S16.1B6.72 GiB1.69 GiB9.25 GiB0.05 GiB63±12.9%
gemma-3n-E2B-itF165.4B8.31 GiB0.14 GiB9.25 GiB0.05 GiB63±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 3080 run?
1535 of 2118 indexed open-weight models fit a GeForce RTX 3080 at 8,192 context with f16 KV cache, the largest being Snowpiercer-15B-v4-heretic at I1-Q3_K_M. That covers text, vision-language, image, video and speech models.
How much usable memory does a GeForce RTX 3080 actually have?
Its nameplate is 10 GB, but about 9.30 GiB is available to a model once driver and compositor overhead is accounted for.
Is a GeForce RTX 3080 fast for local AI?
Its memory bandwidth is 760 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.
GeForce RTX 3080 — what AI models can it run locally? — ossmodeldb