Desktop GPU · NVIDIA
What AI models can a GeForce RTX 3080 10GB run?
10 GB is the point where local models stop being a demo. The popular 7B to 14B models fit with room for a real conversation, and a 32B fits if you accept a lower quantisation. At 760 GB/s there is enough bandwidth to keep generation responsive for anything that fits.
The short answer
Assuming an 8k context window and default settings, these are the models worth downloading first.
Best all-rounder
Gemma 4 12B Instruct
Q5_K_M · 9.13 GB · about 75.1 tokens/s
Best for code
DeepSeek Coder 6.7B Instruct
Q4_K_M · 8.64 GB · about 80 tokens/s
Largest that still runs well
DeepSeek Coder V2 Lite Instruct
15.71B parameters · IQ4_XS · about 391.2 tokens/s
Fastest useful answer
PowerMoE 3B
about 454.6 tokens/s · 4.53 GB
See how fast it feels
GeForce RTX 3080 10GB running Gemma 4 12B Instruct at Q5_K_M
YouWhy does my model use more memory when the conversation gets longer?
Because of the KV cache. Every token you send leaves behind a key and a value vector in each layer of the model, and those stay in memory for as long as the conversation lasts.
The weights are a fixed cost: load a 4-bit 8B model and that is about 4.8 GB, whether you write one word or ten thousand. The cache is the part that grows, and it grows in a straight line with the number of tokens in the window.
How steeply depends on the model's attention design. With grouped-query attention, several query heads share one key-value pair, which cuts the cache by that ratio. Without it, every head keeps its own, and a long context can cost more memory than the weights themselves.
If you are short on memory, the first thing to try is lowering the context window in your runtime, not the quantisation.
Reading your question
Simulated from our estimate at a 512-token question, not a recording. Assumes nothing else is competing for the GPU.
Every model, scored on this device
Each row uses the highest-quality quantisation that both fits and stays conversational. Speed is a single-stream estimate at 8k context.
| Model | Params | Verdict | Download | Memory used | Speed | Max context |
|---|---|---|---|---|---|---|
| Gemma 4 12B Instruct Gemma | 11.96B | Just fits | Q5_K_M | 9.13 GB | 75.1 t/s | 8k |
| DeepSeek Coder V2 Lite Instruct DeepSeek | 15.71B (2.74B active) | Just fits | IQ4_XS | 8.74 GB | 391.2 t/s | 16k |
| vLLM Translategemma 12B Instruct Gemma | 13.19B | Just fits | Q4_K_M | 8.63 GB | 80 t/s | 16k |
| Gemma 2 9B Instruct Gemma | 9.24B | Just fits | Q6_K | 9.19 GB | 74.5 t/s | 8k |
| Mistral Nemo Instruct 2407 Mistral | 12.25B | Just fits | Q4_K_M | 9.05 GB | 76.6 t/s | 8k |
| Gemma 3 12B Instruct Gemma | 12.19B | Just fits | Q4_K_M | 8.5 GB | 81.3 t/s | 8k |
| Granite 4.1 8B Granite | 8.79B | Just fits | Q6_K | 8.81 GB | 78.3 t/s | 8k |
| Fanar 1 9B Instruct Fanar | 8.78B | Just fits | Q6_K | 8.84 GB | 77.7 t/s | 4k |
| Internlm3 8B Instruct InternLM | 8.8B | Runs great | Q6_K | 7.95 GB | 87.8 t/s | 16k |
| Qwen3.5 9B Qwen | 9.65B | Just fits | Q5_K_M | 8.24 GB | 84.3 t/s | 8k |
| LFM2.5 8B A1B Liquid | 8.47B (1.57B active) | Runs great | Q6_K | 7.57 GB | 395.9 t/s | 32k |
| Gemma 4 E4B Instruct Gemma | 8B | Just fits | Q8_0 | 8.72 GB | 78.3 t/s | 32k |
| Qwen2.5 VL 7B Instruct Qwen | 8.29B | Runs great | Q6_K | 7.59 GB | 92.1 t/s | 16k |
| Qwen3 8B Qwen | 8.19B | Just fits | Q6_K | 8.23 GB | 84.4 t/s | 8k |
| Granite 3.0 8B Instruct Granite | 8.17B | Just fits | Q6_K | 8.34 GB | 83.2 t/s | 4k |
| T Lite Instruct 2.1 T-Lite | 8.19B | Just fits | Q6_K | 8.23 GB | 84.4 t/s | 8k |
| Qwen2.5 7B Instruct Qwen | 7.62B | Just fits | Q8_0 | 8.8 GB | 78.1 t/s | 8k |
| Llama 3.1 8B Instruct Llama | 8.03B | Runs great | Q6_K | 7.98 GB | 87.4 t/s | 8k |
| Apertus 8B Instruct 2509 Apertus | 8.05B | Runs great | Q6_K | 8 GB | 87.2 t/s | 8k |
| Llama 3 Taiwan 8B Instruct Llama | 8.03B | Runs great | Q6_K | 7.98 GB | 87.4 t/s | 8k |
| Mistral 7B Instruct V0.3 Mistral | 7.25B | Just fits | Q8_0 | 9.02 GB | 76.2 t/s | 8k |
| Mistral 7B Instruct V0.2 Mistral | 7.24B | Just fits | Q8_0 | 9.01 GB | 76.3 t/s | 8k |
| Qwen2.5 14B Instruct Qwen | 14.77B | Just fits | Q3_K_M | 9.14 GB | 75.8 t/s | 8k |
| Qwen3 14B Qwen | 14.77B | Just fits | Q3_K_M | 8.89 GB | 78.2 t/s | 8k |
| Phi 4 Phi | 14.66B | Just fits | Q3_K_M | 9.15 GB | 75.7 t/s | 8k |
| OLMo 3 7B Instruct OLMo | 7.3B | Just fits | Q6_K | 8.43 GB | 82.3 t/s | 32k |
| OLMoE 1B 7B 0125 Instruct OLMo | 6.92B (1.28B active) | Just fits | Q8_0 | 8.57 GB | 274.9 t/s | 4k |
| Qwen1.5 MoE A2.7B Qwen | 14.32B (2.69B active) | Just fits | Q3_K_M | 8.74 GB | 228.7 t/s | 8k |
| GPT OSS 20B GPT-OSS | 20.91B (4.18B active) | Just fits | IQ3_XXS ! | 8.23 GB | 416.9 t/s | 32k |
| DeepSeek Coder 7B Instruct V1.5 DeepSeek | 6.91B | Just fits | Q5_K_M | 9.18 GB | 74.8 t/s | 4k |
| Qwen1.5 7B Qwen | 7.72B | Just fits | Q4_K_M | 9.19 GB | 74.7 t/s | 8k |
| Gemma 4 E2B Instruct Gemma | 5.12B | Runs great | Q8_0 | 5.78 GB | 122.6 t/s | 128k |
| CodeLlama 7B Llama | 6.74B | Just fits | Q4_K_M | 8.64 GB | 80 t/s | 8k |
| DeepSeek Coder 6.7B Instruct DeepSeek | 6.74B | Just fits | Q4_K_M | 8.64 GB | 80 t/s | 8k |
| Falcon 7B Falcon | 7.22B | Just fits | IQ4_XS | 8.89 GB | 77.8 t/s | 8k |
| Qwen3.5 4B Qwen | 4.66B | Runs great | Q8_0 | 6.37 GB | 111.1 t/s | 16k |
| Agents A1 4B Other | 4.54B | Runs great | Q8_0 | 6.25 GB | 113.5 t/s | 16k |
| Gemma 3 4B Instruct Gemma | 4.3B | Runs great | Q8_0 | 5.31 GB | 136.9 t/s | 64k |
| Phi 3 Vision 128k Instruct Phi | 4.15B | Runs great | Q8_0 | 7.89 GB | 87.7 t/s | 8k |
| Qwen3 4B Qwen | 4.02B | Runs great | Q8_0 | 5.86 GB | 122.1 t/s | 16k |
| Phi 4 Mini Instruct Phi | 3.84B | Runs great | Q8_0 | 5.59 GB | 129.8 t/s | 32k |
| Phi 3 Mini 4k Instruct Phi | 3.82B | Runs great | Q8_0 | 5.32 GB | 137.6 t/s | 4k |
| PowerLM 3B PowerLM | 3.51B | Runs great | Q8_0 | 7.03 GB | 99.1 t/s | 4k |
| Granite 4.1 3B Granite | 3.4B | Runs great | Q8_0 | 4.75 GB | 156.2 t/s | 32k |
| PowerMoE 3B PowerLM | 3.37B (0.88B active) | Runs great | Q8_0 | 4.53 GB | 454.6 t/s | 4k |
| Llama 3.2 3B Instruct Llama | 3.21B | Runs great | Q8_0 | 4.84 GB | 153.8 t/s | 32k |
| Qwen2.5 3B Instruct Qwen | 3.09B | Runs great | Q8_0 | 4.07 GB | 186.6 t/s | 32k |
| SmolLM3 3B Base SmolLM | 3.08B | Runs great | Q8_0 | 4.34 GB | 172.6 t/s | 64k |
| Starcoder2 3B StarCoder | 3.03B | Runs great | Q8_0 | 3.9 GB | 200 t/s | 16k |
| GLM 4.7 Flash GLM | 31.22B (3.66B active) | Just fits | IQ2_XXS ! | 8.63 GB | 482.8 t/s | 16k |
| Qwen3 30B A3B Qwen | 30.53B (3.34B active) | Just fits | IQ2_XXS ! | 8.8 GB | 401.8 t/s | 8k |
| Phi 2 Phi | 2.78B | Runs great | Q8_0 | 6.01 GB | 118.7 t/s | 2k |
| Gemma 3 27B Instruct Gemma | 27.43B | Just fits | IQ2_XXS ! | 8.59 GB | 81.3 t/s | 8k |
| LFM2.5 2.6B Liquid | 2.7B | Runs great | Q8_0 | 3.87 GB | 198.4 t/s | 64k |
| Gemma 4 26B A4B Instruct Gemma | 25.81B | Runs great | IQ2_XXS ! | 7.2 GB | 97 t/s | 64k |
| Gemma 2 2B Instruct Gemma | 2.61B | Runs great | Q8_0 | 3.73 GB | 208.5 t/s | 8k |
| Mistral Small 24B Instruct 2501 Mistral | 23.57B | Runs great | IQ2_XXS ! | 7.82 GB | 90.3 t/s | 8k |
| Codestral 22B V0.1 Mistral | 22.25B | Runs great | IQ2_XXS ! | 8.06 GB | 87.9 t/s | 8k |
| Qwen3.5 2B Qwen | 2.27B | Runs great | Q8_0 | 3.35 GB | 237.8 t/s | 64k |
| OneRec 1.7B Other | 2.13B | Runs great | Q8_0 | 3.71 GB | 208.9 t/s | 32k |
| Qwen3 1.7B Qwen | 2.03B | Runs great | Q8_0 | 3.61 GB | 216.1 t/s | 32k |
| DeepSeek R1 Distill Qwen 1.5B Qwen | 1.78B | Runs great | Q8_0 | 2.57 GB | 333.1 t/s | 128k |
| Qwen3 1.7B Base Qwen | 1.72B | Runs great | Q8_0 | 3.3 GB | 241.8 t/s | 32k |
| SmolLM2 1.7B SmolLM | 1.71B | Runs great | Q8_0 | 3.92 GB | 195.2 t/s | 8k |
| Qwen2.5 1.5B Instruct Qwen | 1.54B | Runs great | Q8_0 | 2.44 GB | 357.6 t/s | 32k |
| Pythia 1.4B Pythia | 1.52B | Runs great | Q8_0 | 3.73 GB | 207.5 t/s | 2k |
| OLMo 2 0425 1B OLMo | 1.48B | Runs great | Q8_0 | 3.19 GB | 252.9 t/s | 4k |
| Llama 3.2 1B Instruct Llama | 1.24B | Runs great | Q8_0 | 2.2 GB | 421.9 t/s | 128k |
| LFM2.5 1.2B Instruct Liquid | 1.17B | Runs great | Q8_0 | 2.13 GB | 442.7 t/s | 64k |
| TinyLlama 1.1B Chat V1.0 Llama | 1.1B | Runs great | Q8_0 | 1.99 GB | 494.5 t/s | 2k |
| MiniCPM5 1B MiniCPM | 1.08B | Runs great | Q8_0 | 1.95 GB | 496.1 t/s | 128k |
| Gemma 3 1B Instruct Gemma | 1B | Runs great | Q8_0 | 1.71 GB | 599.9 t/s | 32k |
| Qwen3.5 0.8B Qwen | 0.87B | Runs great | Q8_0 | 1.9 GB | 504.3 t/s | 128k |
| Sarashina2.2 0.5B Instruct V0.1 Sarashina | 0.79B | Runs great | Q8_0 | 1.93 GB | 498.4 t/s | 8k |
| Qwen3 0.6B Qwen | 0.75B | Runs great | Q8_0 | 2.28 GB | 385.4 t/s | 32k |
| Qwen1.5 0.5B Chat Qwen | 0.62B | Runs great | Q8_0 | 2.03 GB | 457.1 t/s | 32k |
| Qwen3 0.6B Base Qwen | 0.6B | Runs great | Q8_0 | 2.13 GB | 424.3 t/s | 32k |
| Pythia 410m Pythia | 0.51B | Runs great | Q8_0 | 1.92 GB | 496.7 t/s | 2k |
| H2o Danube3 500m Chat Danube | 0.51B | Runs great | Q8_0 | 1.57 GB | 708.5 t/s | 8k |
| Qwen2.5 0.5B Instruct Qwen | 0.49B | Runs great | Q8_0 | 1.23 GB | 1077 t/s | 32k |
| SmolLM2 360M SmolLM | 0.36B | Runs great | Q8_0 | 1.33 GB | 931.9 t/s | 8k |
| LFM2.5 350M Liquid | 0.35B | Runs great | Q8_0 | 1.26 GB | 1045 t/s | 64k |
| Pythia 160m Pythia | 0.21B | Runs great | Q8_0 | 1.14 GB | 1274.3 t/s | 2k |
| Japanese GPT NeoX Small GPT-NeoX | 0.2B | Runs great | Q8_0 | 1.13 GB | 1300.6 t/s | 2k |
| Llama 160m Llama | 0.16B | Runs great | Q8_0 | 1.09 GB | 1417.7 t/s | 2k |
| LLM Jp 3 150m LLM-jp | 0.15B | Runs great | Q8_0 | 0.97 GB | 1855.2 t/s | 4k |
| SmolLM2 135M SmolLM | 0.13B | Runs great | Q8_0 | 0.94 GB | 2047.2 t/s | 8k |
| Pythia 70m Deduped Pythia | 0.1B | Runs great | Q8_0 | 0.82 GB | 3234 t/s | 2k |
| Qwen3.5 35B A3B Qwen | 35.95B (2.9B active) | Partial offload 16/40 layers | — | 21.57 GB | 22.6 t/s | — |
| Qwen3 Next 80B A3B Instruct Qwen | 81.32B (3.19B active) | Partial offload 8/48 layers | — | 47.2 GB | 14.8 t/s | — |
Showing the 90 best results of 133. The remaining 43 need more memory than this device has, at any quantisation.
What it will not run
4 of the 133 architectures we track are out of reach here, even at two-bit precision. Another 41 run by splitting layers between the card and system RAM, which works but drops generation to single digits.
If you want more headroom
The next steps up in memory, in order. More memory changes which models load at all; more bandwidth changes how fast they answer, so the two are worth weighing separately.
| Device | Memory | Bandwidth | Models that fit | |
|---|---|---|---|---|
| GeForce RTX 2080 Ti | 11 GB | 616 GB/s | 93 | Check price |
| GeForce GTX 1080 Ti | 11 GB | 484 GB/s | 93 | Check price |
| GeForce RTX 3080 Ti | 12 GB | 912 GB/s | 96 | Check price |
| GeForce RTX 3080 12GB | 12 GB | 912 GB/s | 96 | Check price |
Price links go to an Amazon search for the model name and are affiliate links: if you buy through one, we earn a commission at no cost to you. We do not take payment for placement, and the ordering above is by memory capacity alone.
Other devices with 10 GB
Same capacity, different speed. Once a model fits, bandwidth is what separates these.