NVIDIA · consumer

GeForce RTX 2070

GeForce RTX 2070 has 8 GB of VRAM at 448 GB/s — about 7.44 GiB usable after driver and compositor overhead. 1421 of 2118 indexed models fit at 8K context with q4_0 KV.

Spec sheet· bandwidth, theoreticalFrom the file· fit from summed bytesPredicted· speed
Memory
8 GB
GDDR6
Bandwidth
448 GB/s
256-bit bus
Tensor FP16
60 TF
dense
TDP
175 W
$499 MSRP
Context4K8K16K32K64K128K
KV cachef16q8_0q4_0quantizing the KV cache is a ~2× lever on the dominant term at long context
text 1223vision language 103video 8embedding 26audio tts 21image 2audio asr 38

What fits at 8K context

largest quantization that fits, per model · 1421 of 2118 indexed
ModelBest quantParamsWeights●KV●Total in memory◐Headroom◐tok/s◐
spoomplesmaxx-mini-14BI1-IQ3_S14.8B6.23 GiB0.35 GiB7.44 GiB0.00 GiB48±12.9%
vanilla-cn-roleplay-0.2I1-IQ3_S14.8B6.23 GiB0.35 GiB7.44 GiB0.00 GiB48±12.9%
Claria-14bI1-IQ3_S14.8B6.23 GiB0.35 GiB7.44 GiB0.00 GiB48±12.9%
NTX-2.1-ProI1-IQ3_S14.8B6.23 GiB0.35 GiB7.44 GiB0.00 GiB48±12.9%
Qwen3-14B-UncensoredI1-IQ3_S14.8B6.23 GiB0.35 GiB7.44 GiB0.00 GiB48±12.9%
FrogMini-14B-2510I1-IQ3_S—6.23 GiB0.35 GiB7.44 GiB0.00 GiB48±12.9%
Qwen3-14B-abliteratedI1-IQ3_S14.8B6.23 GiB0.35 GiB7.44 GiB0.00 GiB48±12.9%
Hermes-4-14BIQ3_S14.8B6.23 GiB0.35 GiB7.44 GiB0.00 GiB48±12.9%
Slava-Qwen3-14B-SerbianI1-IQ3_S14.8B6.23 GiB0.35 GiB7.44 GiB0.00 GiB48±12.9%
Huihui-Qwen3-14B-abliterated-v2I1-IQ3_S14.8B6.23 GiB0.35 GiB7.44 GiB0.00 GiB48±12.9%
OmniAtlas-Qwen3-30B-A3BI1-IQ1_M31.7B6.59 GiB0.00 GiB7.44 GiB0.00 GiB48±12.9%
Qwen3-Omni-30B-A3B-CaptionerI1-IQ1_M31.7B6.59 GiB0.00 GiB7.44 GiB0.00 GiB48±12.9%
Ministral-3-3B-Instruct-2512BF163.8B6.40 GiB0.23 GiB7.44 GiB0.00 GiB48±12.9%
Ministral-3-3B-Reasoning-2512BF164.3B6.40 GiB0.23 GiB7.44 GiB0.00 GiB48±12.9%
Ministral-3-3B-Instruct-2512-BF16BF164.3B6.39 GiB0.23 GiB7.44 GiB0.00 GiB48±12.9%
Amaretto-3BF164.3B6.39 GiB0.23 GiB7.44 GiB0.00 GiB48±12.9%
Grug-12BIQ4_XS12.0B6.32 GiB0.27 GiB7.43 GiB0.01 GiB48±12.9%
gemma-4-12B-it-Esper4IQ4_XS12.0B6.32 GiB0.27 GiB7.43 GiB0.01 GiB48±12.9%
gemma-4-12B-itIQ4_XS12.0B6.32 GiB0.27 GiB7.43 GiB0.01 GiB48±12.9%
GLM-4.7-Flash-REAP-23B-A3B-absolute-heresyMoEI1-IQ2_S23.0B6.51 GiB0.12 GiB7.43 GiB0.01 GiB162±37%
HomunculusQ3_K_L12.5B6.23 GiB0.35 GiB7.43 GiB0.01 GiB48±12.9%
zeta-2.1I1-Q6_K8.3B6.31 GiB0.28 GiB7.43 GiB0.01 GiB48±12.9%
Ministral-3-14B-Instruct-2512-BF16-abliteratedI1-Q3_K_M13.9B6.22 GiB0.35 GiB7.43 GiB0.01 GiB48±12.9%
Ministral-3-14B-abliteratedQ3_K_M13.9B6.22 GiB0.35 GiB7.43 GiB0.01 GiB48±12.9%
Ministral-3-14B-Instruct-2512-BF16Q3_K_M13.9B6.22 GiB0.35 GiB7.43 GiB0.01 GiB48±12.9%
Ministral-3-14B-Instruct-2512Q3_K_M13.9B6.22 GiB0.35 GiB7.43 GiB0.01 GiB48±12.9%
Ministral-3-14B-Reasoning-2512-UncensoredI1-Q3_K_M13.9B6.22 GiB0.35 GiB7.43 GiB0.01 GiB48±12.9%
Ministral-3-14B-Reasoning-2512Q3_K_M13.9B6.22 GiB0.35 GiB7.43 GiB0.01 GiB48±12.9%
gemma-7bI1-Q5_K_S8.5B5.57 GiB0.98 GiB7.43 GiB0.01 GiB48±12.9%
granite-3.3-8b-instructQ6_K8.2B6.24 GiB0.35 GiB7.43 GiB0.01 GiB48±12.9%
granite-3.1-8b-instructQ6_K_M8.2B6.24 GiB0.35 GiB7.43 GiB0.01 GiB48±12.9%
granite-3.2-8b-instructQ6_K8.2B6.24 GiB0.35 GiB7.43 GiB0.01 GiB48±12.9%
Qwen3-VL-8B-Instruct-HereticI1-IQ3_XXS8.8B6.28 GiB0.32 GiB7.43 GiB0.01 GiB48±12.9%
LFM2.5-8B-A1BMoEUD-Q6_K8.5B6.60 GiB0.03 GiB7.42 GiB0.02 GiB140±37%
glm-4-9b-chat-1mIQ4_NL9.5B5.17 GiB1.41 GiB7.42 GiB0.02 GiB48±12.9%
MythoMax-L2-13bI1-IQ3_XXS13.0B4.82 GiB1.76 GiB7.42 GiB0.02 GiB48±12.9%
Fimbulvetr-11B-v2I1-Q4_K_M10.7B6.16 GiB0.42 GiB7.42 GiB0.02 GiB48±12.9%
gemma-4-12B-it-hereticQ4_K_S12.0B6.30 GiB0.27 GiB7.42 GiB0.02 GiB48±12.9%
NVIDIA-Nemotron-Nano-9B-v2Q4_K_M8.9B6.08 GiB0.49 GiB7.42 GiB0.02 GiB48±12.9%
openNemo-9B-abliteratedQ4_K_M8.9B6.08 GiB0.49 GiB7.42 GiB0.02 GiB48±12.9%
v6-Finch-7B-HFQ5_K_L7.6B5.45 GiB1.13 GiB7.42 GiB0.02 GiB48±12.9%
rwkv-6-world-7bQ5_K_L7.6B5.45 GiB1.13 GiB7.42 GiB0.02 GiB48±12.9%
Tess-4-9BQ5_K_M9.7B6.51 GiB0.07 GiB7.42 GiB0.02 GiB48±12.9%
gemma-4-E4B-it-hereticQ6_K8.0B6.55 GiB0.05 GiB7.41 GiB0.03 GiB48±12.9%
next-8bI1-Q6_K8.2B6.26 GiB0.32 GiB7.41 GiB0.03 GiB48±12.9%
Supertron2-Reranker-8BI1-Q6_K8.8B6.26 GiB0.32 GiB7.41 GiB0.03 GiB48±12.9%
next-ocrI1-Q6_K8.8B6.26 GiB0.32 GiB7.41 GiB0.03 GiB48±12.9%
Qwen3-VL-8B-GLM-4.7-Flash-Heretic-Uncensored-ThinkingI1-Q6_K8.8B6.26 GiB0.32 GiB7.41 GiB0.03 GiB48±12.9%
Midas-FableAgent-8BI1-Q6_K8.2B6.26 GiB0.32 GiB7.41 GiB0.03 GiB48±12.9%
Qwen3-VL-8B-Heretic-1.3.0I1-Q6_K8.8B6.26 GiB0.32 GiB7.41 GiB0.03 GiB48±12.9%
Qwen3-VL-8B-Thinking-Unredacted-MAXI1-Q6_K8.8B6.26 GiB0.32 GiB7.41 GiB0.03 GiB48±12.9%
Qwen3-VL-8B-Instruct-Minecraft-MT-en-zhI1-Q6_K8.8B6.26 GiB0.32 GiB7.41 GiB0.03 GiB48±12.9%
Qwen-3-VL-8B-Instruct-hereticI1-Q6_K8.8B6.26 GiB0.32 GiB7.41 GiB0.03 GiB48±12.9%
Poe-8B-GLM5-Opus4.6-Sonnet4.5-Kimi-Grok-Gemini-3-pro-preview-HERETICI1-Q6_K8.8B6.26 GiB0.32 GiB7.41 GiB0.03 GiB48±12.9%
ToolCUA-8BI1-Q6_K8.8B6.26 GiB0.32 GiB7.41 GiB0.03 GiB48±12.9%
Qwen3-VL-8B-ThinkingQ6_K8.8B6.26 GiB0.32 GiB7.41 GiB0.03 GiB48±12.9%
Huihui-Qwen3-VL-8B-Instruct-abliteratedI1-Q6_K8.8B6.26 GiB0.32 GiB7.41 GiB0.03 GiB48±12.9%
Qwen3-VL-8B-Instruct-Unredacted-MAXQ6_K8.8B6.26 GiB0.32 GiB7.41 GiB0.03 GiB48±12.9%
Qwen3-VL-Reranker-8BI1-Q6_K8.8B6.26 GiB0.32 GiB7.41 GiB0.03 GiB48±12.9%
Salience-1-9BI1-Q6_K8.8B6.26 GiB0.32 GiB7.41 GiB0.03 GiB48±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.

Measured on this card

third-party benchmarks, aggregated
Workload◍MedianMiddle 50%Runs
Image generation5.75 it/s4.02–6.72293
Benchmarked· n=293

Aggregated from community-submitted runs, so the spread is wide by nature — it covers different models, resolutions, step counts and settings, not one controlled configuration. Read the middle 50% rather than the median alone. These figures are reproduced with attribution from vladmandic-sd-data-benchmark, which publishes no licence — so we display and link rather than redistribute them.

Questions people ask

What AI models can a GeForce RTX 2070 run?
1421 of 2118 indexed open-weight models fit a GeForce RTX 2070 at 8,192 context with q4_0 KV cache, the largest being spoomplesmaxx-mini-14B at I1-IQ3_S. That covers text, vision-language, image, video and speech models.
How much usable memory does a GeForce RTX 2070 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 2070 fast for local AI?
Its memory bandwidth is 448 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.