The RTX 3080 10GB has high bandwidth (760 GB/s) but only 10 GB of GDDR6X. That makes it very fast for the models that fit, and awkward for the ones that do not.
Check your own setup with the Can My GPU Run This AI Model? calculator. The numbers below come from the same data.
RTX 3080 10GB specs that matter for local AI
- Memory: 10 GB GDDR6X on a 320-bit bus
- Memory bandwidth: 760 GB/s
- Note: a 12 GB version of the RTX 3080 also exists; this page covers the 10 GB model.
VRAM decides which models fit. Bandwidth decides how fast they generate text, because every new token means reading the whole model from memory.
Which AI models fit on the RTX 3080 10GB?
The table assumes Q4_K_M quantization (about 4.9 bits per weight) and an 8K context. Totals include the KV cache and about 1 GB of runtime overhead. Speed is an estimate: memory bandwidth divided by model size, at 60 to 80 percent efficiency, which matches published llama.cpp results on several RTX cards.
| Model | Weights | KV cache (8K) | Total (est.) | On 10 GB | Speed (est.) |
|---|---|---|---|---|---|
| Qwen3 4B | 2.5 GB | 1.2 GB | 4.7 GB | Fits | 185 to 247 tok/s |
| Llama 3.1 8B Instruct | 4.9 GB | 1.1 GB | 7.0 GB | Fits | 93 to 124 tok/s |
| Qwen3 8B | 5.0 GB | 1.2 GB | 7.2 GB | Fits | 91 to 121 tok/s |
| Gemma 3 12B Instruct | 7.5 GB | 0.9 GB | 9.3 GB | Tight | 61 to 81 tok/s |
| Qwen3 14B | 9.0 GB | 1.3 GB | 11.4 GB | No | – |
| OpenAI gpt-oss-20b | 12.8 GB | 0.2 GB | 14.0 GB | No | – |
| Mistral Small 3.2 24B | 14.7 GB | 1.3 GB | 17.0 GB | No | – |
| Qwen3.6 27B | 17.0 GB | 0.5 GB | 18.6 GB | No | – |
| Gemma 3 27B Instruct | 16.8 GB | 1.1 GB | 18.9 GB | No | – |
| Qwen3 32B | 20.1 GB | 2.1 GB | 23.2 GB | No | – |
| Qwen3.6 35B-A3B | 22.0 GB | 0.2 GB | 23.2 GB | No | – |
“Fits” means the estimate is at most 92 percent of VRAM. “Tight” means it is between 92 and 100 percent, so close other GPU apps. Mixture-of-Experts models (gpt-oss-20b and Qwen3.6 35B-A3B) need their full weights in memory, but they only use part of them per token, so speed is not estimated.
Best models for the RTX 3080 10GB
- Llama 3.1 8B and Qwen3 8B: about 7 GB at Q4; Q6_K also fits (about 8.7 GB at 8K).
- Qwen3 4B: very fast with lots of room for context.
- Gemma 3 12B (Q4_K_M): about 9.3 GB at 8K, which is tight on 10 GB.
How context length changes things
The KV cache grows with every token of context, so a model that fits at 4K can fail at 32K.
| Model (Q4_K_M) | Context | KV cache | Total (est.) | On 10 GB |
|---|---|---|---|---|
| Llama 3.1 8B Instruct | 4K | 0.54 GB | 6.5 GB | Fits |
| Llama 3.1 8B Instruct | 8K | 1.07 GB | 7.0 GB | Fits |
| Llama 3.1 8B Instruct | 16K | 2.15 GB | 8.1 GB | Fits |
| Llama 3.1 8B Instruct | 32K | 4.29 GB | 10.2 GB | No |
| Qwen3 14B | 4K | 0.67 GB | 10.7 GB | No |
| Qwen3 14B | 8K | 1.34 GB | 11.4 GB | No |
| Qwen3 14B | 16K | 2.68 GB | 12.7 GB | No |
| Qwen3 14B | 32K | 5.37 GB | 15.4 GB | No |
A Q8 KV cache (an option in Ollama and llama.cpp) roughly halves the cache cost if you need more context.
What does not fit
Qwen3 14B (about 11.4 GB), gpt-oss-20b and bigger models do not fit in 10 GB at Q4. The 8B model at Q8_0 (about 10.6 GB at 8K) is also just over.
Speedy, but short on memory. For 8B models it is excellent. For anything bigger, VRAM becomes the problem long before speed does.
Verdict
What speed to expect
A community llama.cpp dataset (May 2024) measured the RTX 3080 10GB at 109.6 tokens per second on Llama 3 8B Q4_K_M. Our bandwidth-based estimate is 93 to 124 tokens per second, so the measured result sits in the middle of the range. FP16 ran out of memory, as expected for 16 GB of weights.
Frequently asked questions
Can the RTX 3080 10GB run a 13B or 14B model?
Not at Q4 with a normal context: Qwen3 14B needs about 11.4 GB.
How fast is it on 8B models?
About 110 tokens per second on Llama 3 8B Q4_K_M in a community dataset.
Is 10 GB enough for AI?
For 8B models, yes. Beyond that, it is limiting.
Sources and method
- RTX 3080 specs: Guru3D’s RTX 3080 and 3090 specs article (10 GB, 760 GB/s).
- Measured speeds: XiongjieDai’s GPU-Benchmarks-on-LLM-Inference on GitHub (llama.cpp, May 2024 snapshot).
- Linear-attention and sliding-window layers are handled with reduced KV cache; small fixed states in some newer models are not counted.
- Model architecture: each model’s published
config.jsonon Hugging Face (Llama 3.1 8B and Gemma 3 are read from the unsloth mirrors because the original repositories require a login). - VRAM and speed figures are estimates from our calculator, not our own benchmarks. Last updated 3 Nov 2026.
