VRAM guide

Best local LLMs for 16GB VRAM

16GB covers the RTX 4060 Ti 16GB, RTX 5060 Ti 16GB, RTX 4080, and RX 7800 XT. Ranked for GeForce RTX 4060 Ti 16GB at 16 GB. Updated 2026-10-17. These are estimates for llama.cpp with FP16 KV cache, not measured benchmarks. Method.

Model	Size	Verdict	Quant	tok/s
Qwen3.8 27B	27.3B	Tight fit	UD-IQ4_XS	11–15
Devstral Small 2 24B Instruct 2512	23.6B	Tight fit	Q4_K_M	11–14
Mistral Small 3.2 24B Instruct 2506	23.6B	Tight fit	Q4_K_M	11–14
Magistral Small 2509	23.6B	Tight fit	Q4_K_M	11–14
gpt oss 20b	20.9B	Tight fit	UD-Q8_K_XL	39–54
Qwen3.6 27B	26.9B	Tight fit	UD-Q3_K_XL	11–15
Qwen3.5 27B	26.9B	Tight fit	UD-Q3_K_XL	11–15
GLM 4.7 Flash REAP 23B A3B	23.0B	Tight fit	UD-Q4_K_XL	69–94
GLM 4 32B 0414	32.6B	Tight fit	Q3_K_S	11–15
Qwen2.5 14B Instruct	14.8B	Runs	Q5_K_M	13–18
NousResearch Hermes 4 14B	14.8B	Runs	Q5_K_M	13–18
DeepSeek R1 Distill Qwen 14B	14.8B	Tight fit	Q6_K	12–16
Qwen2.5 Coder 14B Instruct 128K	14.8B	Tight fit	Q6_K	12–16
Qwen3 14B	14.8B	Tight fit	UD-Q6_K_XL	11–15
phi 4	14.7B	Tight fit	Q6_K	12–16
gemma 4 26B A4B it	25.2B	Tight fit	UD-IQ4_XS	38–52
Qwen3 32B	32.8B	Tight fit	UD-IQ3_XXS	11–15
Qwen3 VL 32B Instruct	32.8B	Tight fit	UD-IQ3_XXS	11–15
GLM 4.7 Flash	29.9B	Tight fit	UD-Q3_K_XL	47–64
Qwen3.5 35B A3B	34.7B	Tight fit	UD-IQ3_S	118–161
Qwen3.6 35B A3B	34.7B	Tight fit	UD-IQ3_S	117–160
Ministral 3 14B Instruct 2512	13.5B	Tight fit	UD-Q6_K_XL	12–16
Ministral 3 14B Reasoning 2512	13.5B	Tight fit	UD-Q6_K_XL	12–16
Qwen3 Coder 30B A3B Instruct	30.5B	Tight fit	UD-Q3_K_XL	73–100
Qwen3 30B A3B Instruct 2507	30.5B	Tight fit	UD-Q3_K_XL	73–100
Qwen3 VL 30B A3B Instruct	30.5B	Tight fit	UD-Q3_K_XL	73–100
Qwen3 30B A3B Thinking 2507	30.5B	Tight fit	UD-Q3_K_XL	73–100
Qwen2.5 Coder 32B Instruct	32.8B	Tight fit	IQ3_XXS	11–14
LFM2.5 24B A2B	24.0B	Runs	UD-Q3_K_XL	100–136
Kwaipilot KAT Coder V2.5 Dev	34.7B	Tight fit	IQ3_XXS	57–78

92 models fit fully on this reference profile. 128-bit GDDR6 at 18 Gbps = 288 GB/s.