Which LLMs run on 16 GB of VRAM?

25 of our 61 catalog models fit a 16 GB card at Q4/8K. Fastest on the RTX 5080 16GB: Qwen3-0.6B, gemma-3-1b-it, Llama-3.2-1B-Instruct.

What is the largest model that fits in 16 GB VRAM?

gpt-oss-20b at 21.5B — it needs about 13.55 GB at Q4/8K and runs at roughly 204.9 tok/s on the RTX 5080 16GB (measured).

Best LLMs for 16 GB VRAM GPUs (Q4/8K)

Our catalog lists 13 GPUs with 16 GB VRAM (RTX 5080 16GB, RTX 5070 Ti 16GB, RTX 4080 Super 16GB, RTX 4070 Ti Super 16GB, RX 9070 16GB and 8 more). The ranking uses the RTX 5080 16GB — the highest-bandwidth 16 GB card at 960 GB/s — and shows the 25 of 61 catalog models that fully fit it at Q4 / 8K context (weights + KV cache + 1 GB overhead within 95% of VRAM), fastest first. Methodology.

#	Model	Params (B)	Q4 tok/s	VRAM needed (GB)	Fit check
1	Qwen3-0.6B	0.8	763.6 est.	1.6	fit check
2	gemma-3-1b-it	1.0	610.9 est.	1.75	fit check
3	Llama-3.2-1B-Instruct	1.2	509.1 est.	1.9	fit check
4	Qwen3-1.7B	2.0	305.5 est.	2.5	fit check
5	gpt-oss-20b	21.5	204.9 measured	13.55	fit check
6	Llama-3.2-3B-Instruct	3.2	190.9 est.	3.4	fit check
7	Mistral-7B-Instruct-v0.3	7.2	163.6 measured	6.4	fit check
8	Phi-4-mini-instruct	3.8	160.8 est.	3.85	fit check
9	Phi-4-mini-reasoning	3.8	160.8 est.	3.85	fit check
10	gemma-3-4b-it	4.3	156.6 est.	4.15	fit check
11	Qwen3-4B-Instruct-2507	4.0	152.7 est.	4.0	fit check
12	Qwen3-4B-Thinking-2507	4.0	152.7 est.	4.0	fit check
13	Llama-3.1-8B-Instruct	8.0	136.2 measured	7.0	fit check
14	Qwen2.5-7B-Instruct	7.6	80.4 est.	6.7	fit check
15	Qwen2.5-Coder-7B-Instruct	7.6	80.4 est.	6.7	fit check
16	DeepSeek-R1-Distill-Qwen-7B	7.6	80.4 est.	6.7	fit check
17	Qwen3-8B	8.2	74.5 est.	7.15	fit check
18	DeepSeek-R1-0528-Qwen3-8B	8.2	74.5 est.	7.15	fit check
19	gemma-3-12b-it	12.2	50.9 est.	10.11	fit check
20	Phi-4	14.7	41.6 est.	12.03	fit check
21	Phi-4-reasoning	14.7	41.6 est.	12.03	fit check
22	Qwen3-14B	14.8	41.3 est.	12.1	fit check
23	Qwen2.5-14B-Instruct	14.8	41.3 est.	12.1	fit check
24	Qwen2.5-Coder-14B-Instruct	14.8	41.3 est.	12.1	fit check
25	DeepSeek-R1-Distill-Qwen-14B	14.8	41.3 est.	12.1	fit check

tok/s are bandwidth-based estimates unless tagged measured. Every 16 GB card fits the same set of models and ranks them identically; absolute tok/s scales with each card's bandwidth.

Buy the RTX 5080 16GB: NVIDIA GeForce RTX 5080 Founders Edition (affiliate) · More listings — Amazon prices change; the price after you click is the live price.

More lists: 7B-class models · 13B-class models · 30B-class models · 70B+ models · Under $500 · Under $1,000 · Under $2,000 · 8 GB VRAM · 12 GB VRAM · 24 GB VRAM · For Qwen3 · For DeepSeek · For Llama · For Gemma · For GLM · For coding · For reasoning · MoE picks

Best models for 16 GB VRAM