RTX 3080 10GB — local LLM fit & speed

NVIDIA · 10 GB GDDR6X · 760 GB/s · 320 W · MSRP $699 · Search Amazon (affiliate)

The RTX 3080 10GB is a NVIDIA accelerator with 10 GB GDDR6X and about 760 GB/s peak memory bandwidth. AI Hashrate estimates local LLM decode (batch 1) at 8K context: fit counts weights + KV cache + 1 GB runtime overhead — not weights alone. At Q4, about 17 curated models fit fully on this card; at FP16, about 7 fit. Top Q4 speeds: Qwen3-0.6B ≈ 604.5 tok/s (estimated, fits); gemma-3-1b-it ≈ 483.6 tok/s (estimated, fits); Llama-3.2-1B-Instruct ≈ 403.0 tok/s (estimated, fits); Qwen3-1.7B ≈ 241.8 tok/s (estimated, fits); Llama-3.2-3B-Instruct ≈ 151.1 tok/s (estimated, fits). Relative ranking is more reliable than absolute tok/s. See methodology for the bandwidth formula and measured-anchor policy. Methodology.

Table: decode speed estimates (Q4 / FP16). Measured rows override estimates. MSRP is list price, not live retail. Context default 8K.

Model	Params (B)	Quant	tok/s	Fits?
Qwen3-0.6B	0.8	Q4	604.5 est.	Yes
gemma-3-1b-it	1.0	Q4	483.6 est.	Yes
Llama-3.2-1B-Instruct	1.2	Q4	403.0 est.	Yes
Qwen3-1.7B	2.0	Q4	241.8 est.	Yes
gpt-oss-20b	21.5	Q4	209.1 est.	No
Qwen3-0.6B	0.8	FP16	166.2 est.	Yes
Llama-3.2-3B-Instruct	3.2	Q4	151.1 est.	Yes
Qwen3-30B-A3B	30.5	Q4	136.3 est.	No
Qwen3-Coder-30B-A3B-Instruct	30.5	Q4	136.3 est.	No
Qwen3-30B-A3B-Instruct-2507	30.5	Q4	136.3 est.	No
gemma-3-1b-it	1.0	FP16	133.0 est.	Yes
Phi-4-mini-instruct	3.8	Q4	127.3 est.	Yes
Phi-4-mini-reasoning	3.8	Q4	127.3 est.	Yes
gemma-3-4b-it	4.3	Q4	124.0 est.	Yes
Qwen3-4B-Instruct-2507	4.0	Q4	120.9 est.	Yes
Qwen3-4B-Thinking-2507	4.0	Q4	120.9 est.	Yes
gemma-3-12b-it	12.2	Q4	112.7 est.	No
Llama-3.2-1B-Instruct	1.2	FP16	110.8 est.	Yes
Qwen3-4B-Instruct-2507	4.0	FP16	98.9 est.	No
Qwen3-4B-Thinking-2507	4.0	FP16	98.9 est.	No
gemma-3-4b-it	4.3	FP16	90.4 est.	No
Mistral-7B-Instruct-v0.3	7.2	Q4	67.2 est.	Yes
Qwen3-1.7B	2.0	FP16	66.5 est.	Yes
Phi-4	14.7	Q4	65.0 est.	No
Phi-4-reasoning	14.7	Q4	65.0 est.	No
Qwen3-14B	14.8	Q4	63.8 est.	No
Qwen2.5-14B-Instruct	14.8	Q4	63.8 est.	No
Qwen2.5-Coder-14B-Instruct	14.8	Q4	63.8 est.	No
DeepSeek-R1-Distill-Qwen-14B	14.8	Q4	63.8 est.	No
Qwen2.5-7B-Instruct	7.6	Q4	63.6 est.	Yes
Qwen2.5-Coder-7B-Instruct	7.6	Q4	63.6 est.	Yes
DeepSeek-R1-Distill-Qwen-7B	7.6	Q4	63.6 est.	Yes
Llama-3.1-8B-Instruct	8.0	Q4	60.5 est.	Yes
Qwen3-8B	8.2	Q4	59.0 est.	Yes
DeepSeek-R1-0528-Qwen3-8B	8.2	Q4	59.0 est.	Yes
Llama-3.2-3B-Instruct	3.2	FP16	41.6 est.	Yes
Phi-4-mini-instruct	3.8	FP16	35.0 est.	Yes
Phi-4-mini-reasoning	3.8	FP16	35.0 est.	Yes
GLM-4.7-Flash	31.2	Q4	27.2 est.	No
Mistral-7B-Instruct-v0.3	7.2	FP16	18.6 est.	No
Devstral-Small-2507	23.6	Q4	16.7 est.	No
Qwen2.5-7B-Instruct	7.6	FP16	15.9 est.	No
Qwen2.5-Coder-7B-Instruct	7.6	FP16	15.9 est.	No
DeepSeek-R1-Distill-Qwen-7B	7.6	FP16	15.9 est.	No
Mistral-Small-3.2-24B-Instruct-2506	24.0	Q4	15.9 est.	No
Magistral-Small-2509	24.0	Q4	15.9 est.	No
Llama-3.1-8B-Instruct	8.0	FP16	13.7 est.	No
Qwen3-8B	8.2	FP16	12.8 est.	No
DeepSeek-R1-0528-Qwen3-8B	8.2	FP16	12.8 est.	No
Mixtral-8x7B-Instruct-v0.1	46.7	Q4	12.5 est.	No
gemma-3-27b-it	27.4	Q4	11.0 est.	No
Qwen3-32B	32.8	Q4	6.4 est.	No
Qwen2.5-32B-Instruct	32.8	Q4	6.4 est.	No
Qwen2.5-Coder-32B-Instruct	32.8	Q4	6.4 est.	No
DeepSeek-R1-Distill-Qwen-32B	32.8	Q4	6.4 est.	No
gemma-3-12b-it	12.2	FP16	4.1 est.	No
Qwen3-14B	14.8	FP16	2.3 est.	No
Qwen2.5-14B-Instruct	14.8	FP16	2.3 est.	No
Qwen2.5-Coder-14B-Instruct	14.8	FP16	2.3 est.	No
DeepSeek-R1-Distill-Qwen-14B	14.8	FP16	2.3 est.	No
Phi-4	14.7	FP16	2.3 est.	No
Phi-4-reasoning	14.7	FP16	2.3 est.	No