NVIDIA · consumer

GeForce RTX 2060

GeForce RTX 2060 has 6 GB of VRAM at 336 GB/s — about 5.58 GiB usable after driver and compositor overhead. 306 of 2118 indexed models fit at 128K context with f16 KV.

Spec sheet· bandwidth, theoreticalFrom the file· fit from summed bytesPredicted· speed
Memory
6 GB
GDDR6
Bandwidth
336 GB/s
192-bit bus
Tensor FP16
52 TF
dense
TDP
160 W
$349 MSRP
Context4K8K16K32K64K128K
KV cachef16q8_0q4_0quantizing the KV cache is a ~2× lever on the dominant term at long context
text 218audio asr 29audio tts 15vision language 28embedding 13video 3

What fits at 128K context

largest quantization that fits, per model · 306 of 2118 indexed
ModelBest quantParamsWeights●KV●Total in memory◐Headroom◐tok/s◐
alduin-4b-it-baseI1-Q3_K_L4.3B2.09 GiB2.67 GiB5.58 GiB0.00 GiB50±12.9%
Dolphin3.0-Llama3.2-1BQ4_11.2B0.77 GiB4.00 GiB5.58 GiB0.00 GiB50±12.9%
Llama-3.2-1B-Instruct-hereticI1-Q4_11.2B0.77 GiB4.00 GiB5.58 GiB0.00 GiB50±12.9%
Llama-3.2-1B-InstructQ4_11.2B0.77 GiB4.00 GiB5.58 GiB0.00 GiB50±12.9%
canary-qwen-2.5bBF162.6B4.73 GiB0.00 GiB5.57 GiB0.01 GiB50±12.9%
EXAONE-Deep-7.8BQ4_K_L7.8B4.73 GiB0.00 GiB5.57 GiB0.01 GiB50±12.9%
EXAONE-3.5-7.8B-InstructQ4_K_L7.8B4.73 GiB0.00 GiB5.57 GiB0.01 GiB50±12.9%
gemma-2bQ8_02.5B2.49 GiB2.25 GiB5.57 GiB0.01 GiB50±12.9%
Qwen3-TTS-12Hz-0.6B-BaseQ4_K_M915M4.72 GiB0.00 GiB5.57 GiB0.01 GiB50±12.9%
VoxCPM2F162.3B4.72 GiB0.00 GiB5.57 GiB0.01 GiB50±12.9%
gemma-3-4b-it-roleplay-tuned-v1I1-Q3_K_L4.3B2.08 GiB2.67 GiB5.57 GiB0.01 GiB50±12.9%
Gemma-3-4B-VL-it-Gemini-Pro-Heretic-Uncensored-ThinkingQ3_K_L4.3B2.08 GiB2.67 GiB5.57 GiB0.01 GiB50±12.9%
gemma-3-4b-it-roleplay-tuned-v2I1-Q3_K_L4.3B2.08 GiB2.67 GiB5.57 GiB0.01 GiB50±12.9%
gemma-3-4b-it-heretic-uncensored-abliterated-ExtremeI1-Q3_K_L4.3B2.08 GiB2.67 GiB5.57 GiB0.01 GiB50±12.9%
medgemma-4b-itQ3_K_L4.3B2.08 GiB2.67 GiB5.57 GiB0.01 GiB50±12.9%
medgemma-1.5-4b-itQ3_K_L4.3B2.08 GiB2.67 GiB5.57 GiB0.01 GiB50±12.9%
gemma-3-4b-it-abliteratedQ3_K_L4.3B2.08 GiB2.67 GiB5.57 GiB0.01 GiB50±12.9%
amoral-gemma3-4B-v1Q3_K_L4.3B2.08 GiB2.67 GiB5.57 GiB0.01 GiB50±12.9%
Gemma3-4B-CodeCenturionI1-Q3_K_L4.3B2.08 GiB2.67 GiB5.57 GiB0.01 GiB50±12.9%
ArrowMint-Gemma3-4B-YUKI-v0.1I1-Q3_K_L4.3B2.08 GiB2.67 GiB5.57 GiB0.01 GiB50±12.9%
gemma-3-4b-itQ3_K_L4.3B2.08 GiB2.67 GiB5.57 GiB0.01 GiB50±12.9%
t5-v1_1-xxlQ2_K4.8B4.72 GiB0.00 GiB5.56 GiB0.02 GiB50±12.9%
Qwen2-1.5BQ2_K1.5B1.26 GiB3.50 GiB5.56 GiB0.02 GiB50±12.9%
OneLLM-Doey-ChatQA-V1-Llama-3.2-1BI1-Q4_K_M1.2B0.75 GiB4.00 GiB5.55 GiB0.03 GiB50±12.9%
Llama-3.2-1BQ4_K_M1.2B0.75 GiB4.00 GiB5.55 GiB0.03 GiB50±12.9%
Llama-3.2-1B-Instruct-UncensoredI1-Q4_K_M1.2B0.75 GiB4.00 GiB5.55 GiB0.03 GiB50±12.9%
Gemma-3-4b-it-Uncensored-DBL-XI1-IQ2_M4.7B1.55 GiB3.19 GiB5.55 GiB0.03 GiB50±12.9%
LFM2-1.2BQ5_01.2B0.75 GiB4.00 GiB5.55 GiB0.03 GiB50±12.9%
SEX_ROLEPLAY-3.2-1BI1-Q3_K_M1.5B0.75 GiB4.00 GiB5.55 GiB0.03 GiB50±12.9%
Llama-3.2-1B-Instruct-abliteratedI1-Q3_K_M1.5B0.75 GiB4.00 GiB5.55 GiB0.03 GiB50±12.9%
Novaciano-3.2-1BI1-Q3_K_M1.5B0.75 GiB4.00 GiB5.55 GiB0.03 GiB50±12.9%
Imp-RPG.System-1BI1-Q3_K_M1.5B0.75 GiB4.00 GiB5.55 GiB0.03 GiB50±12.9%
gemma-4-E2B-itQ4_K_L5.1B3.85 GiB0.90 GiB5.54 GiB0.04 GiB50±12.9%
Dolphin3.0-Qwen2.5-1.5BQ6_K_L1.5B1.24 GiB3.50 GiB5.54 GiB0.04 GiB50±12.9%
Qwen2.5-Math-1.5B-InstructQ6_K_L1.5B1.24 GiB3.50 GiB5.54 GiB0.04 GiB50±12.9%
Qwen2.5-Coder-1.5B-InstructQ6_K_L1.5B1.24 GiB3.50 GiB5.54 GiB0.04 GiB50±12.9%
Qwen2.5-1.5B-InstructQ6_K_L1.5B1.24 GiB3.50 GiB5.54 GiB0.04 GiB50±12.9%
Qwen2-VL-2B-InstructQ6_K_L2.2B1.24 GiB3.50 GiB5.54 GiB0.04 GiB50±12.9%
FableForge-1.5BQ6_K1.5B1.24 GiB3.50 GiB5.54 GiB0.04 GiB50±12.9%
granite-4.0-7B-A1B-Creative-v0.1MoEI1-Q4_K_M6.7B3.77 GiB1.00 GiB5.54 GiB0.04 GiB84±37%
gemma-3n-E4B-itUD-IQ2_M7.8B2.89 GiB1.80 GiB5.53 GiB0.05 GiB51±12.9%
Miril-Drone-2B-1Q4_K_L5.1B3.83 GiB0.90 GiB5.53 GiB0.05 GiB50±12.9%
InternVL3_5-8BQ4_K_M8.5B4.68 GiB0.00 GiB5.53 GiB0.05 GiB51±12.9%
LFM2-700MQ8_0742M0.74 GiB4.00 GiB5.52 GiB0.06 GiB50±12.9%
nomic-embed-text-v1.5F16137M0.26 GiB4.50 GiB5.52 GiB0.06 GiB50±12.9%
nomic-embed-text-v1F16137M0.26 GiB4.50 GiB5.52 GiB0.06 GiB50±12.9%
granite-4.0-h-tinyMoEQ4_K_S6.9B3.75 GiB1.00 GiB5.52 GiB0.06 GiB85±37%
LFM2-8B-A1BMoEIQ3_XS8.3B3.22 GiB1.50 GiB5.51 GiB0.07 GiB65±37%
DeepScaleR-1.5B-PreviewQ4_K_L1.8B1.20 GiB3.50 GiB5.50 GiB0.08 GiB51±12.9%
Qwen2.5-Coder-1.5B-Instruct-abliteratedQ4_K_L1.8B1.20 GiB3.50 GiB5.50 GiB0.08 GiB51±12.9%
Nemotron-Research-Reasoning-Qwen-1.5BQ4_K_L1.8B1.20 GiB3.50 GiB5.50 GiB0.08 GiB51±12.9%
DeepSeek-R1-Distill-Qwen-1.5BQ4_K_L1.8B1.20 GiB3.50 GiB5.50 GiB0.08 GiB51±12.9%
Darwin-4B-ChimeraI1-IQ2_XXS4.0B1.16 GiB3.53 GiB5.50 GiB0.08 GiB51±12.9%
Qwen2.5-1.5B-Instruct-uncensoredQ5_K_M1.8B1.20 GiB3.50 GiB5.50 GiB0.08 GiB51±12.9%
VibeThinker-1.5BQ5_K_M1.8B1.20 GiB3.50 GiB5.50 GiB0.08 GiB51±12.9%
DeepSeek-R1-Distill-Qwen-1.5B-uncensoredI1-Q5_K_M1.8B1.20 GiB3.50 GiB5.50 GiB0.08 GiB51±12.9%
dots.ocrI1-Q5_K_M3.0B1.20 GiB3.50 GiB5.50 GiB0.08 GiB51±12.9%
DeepSeek-R1-Distill-Qwen-1.5B-Fully-UncensoredQ5_K_M1.8B1.20 GiB3.50 GiB5.50 GiB0.08 GiB51±12.9%
jina-embeddings-v5-text-nanoQ8_0212M0.22 GiB4.50 GiB5.49 GiB0.09 GiB50±12.9%
granite-4.0-h-tiny-baseMoEQ4_K_S6.9B3.72 GiB1.00 GiB5.49 GiB0.09 GiB85±37%
●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 2060 run?
306 of 2118 indexed open-weight models fit a GeForce RTX 2060 at 131,072 context with f16 KV cache, the largest being alduin-4b-it-base at I1-Q3_K_L. That covers text, vision-language, image, video and speech models.
How much usable memory does a GeForce RTX 2060 actually have?
Its nameplate is 6 GB, but about 5.58 GiB is available to a model once driver and compositor overhead is accounted for.
Is a GeForce RTX 2060 fast for local AI?
Its memory bandwidth is 336 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.