Hardware

NVIDIA T4

NVIDIA T4 has 16 GB GDDR6 and 320 GB/s. See which open models fit, at which quant, and roughly how fast. Figures below use the default memory (16 GB), 32 GB system RAM, and DDR5-5600 for anything that spills. Updated 2026-10-24.

Memory16 GB GDDR6

Bandwidth320 GB/s

Kindgpu

Year2018

16GB GDDR6, 320 GB/s. Source.

RSS for new models on this device

Models that fit

Model	Verdict	Quant	Context	tok/s
Devstral Small 2 24B Instruct 2512 23.6B	Tight fit	UD-Q4_K_XL	4K	12–16
Mistral Small 3.2 24B Instruct 2506 23.6B	Tight fit	UD-Q4_K_XL	4K	12–16
Magistral Small 2509 23.6B	Tight fit	UD-Q4_K_XL	4K	12–16
Qwen3.8 27B 27.3B	Tight fit	UD-IQ4_XS	8K	12–16
Qwen3.5 27B 26.9B	Tight fit	IQ4_XS	4K	12–16
Swift Qwen3.8 27B 27.3B	Tight fit	Q3_K_L	8K	12–16
gpt oss 20b 20.9B	Runs	UD-Q8_K_XL	8K	44–60
gemma 3 27b it 27.0B	Tight fit	UD-Q3_K_XL	8K	9.9–13
medgemma 27b text it 27.0B	Tight fit	UD-Q3_K_XL	8K	9.9–13
Qwen3.6 27B 26.9B	Tight fit	UD-Q3_K_XL	8K	12–16
LFM2.5 24B A2B 24.0B	Tight fit	UD-Q4_K_XL	4K	112–152
GLM 4.7 Flash REAP 23B A3B 23.0B	Tight fit	UD-Q4_K_XL	8K	76–104
Qwen2.5 32B Instruct 32.8B	Tight fit	Q3_K_S	4K	11–16
Qwen3 32B 32.8B	Tight fit	Q3_K_S	4K	11–16
Qwen2.5 Coder 32B Instruct 32.8B	Tight fit	Q3_K_S	4K	11–16
Qwen3 VL 32B Instruct 32.8B	Tight fit	Q3_K_S	4K	11–16
GLM 4 32B 0414 32.6B	Tight fit	Q3_K_S	8K	12–16
Qwen2.5 14B Instruct 14.8B	Runs	Q5_K_M	8K	15–20
NousResearch Hermes 4 14B 14.8B	Runs	Q5_K_M	8K	15–20
DeepSeek R1 Distill Qwen 14B 14.8B	Tight fit	Q6_K	8K	13–17
Qwen2.5 Coder 14B Instruct 128K 14.8B	Tight fit	Q6_K	8K	13–17
Qwen3 14B 14.8B	Tight fit	UD-Q6_K_XL	8K	12–16
phi 4 14.7B	Tight fit	Q6_K	8K	13–18
gemma 4 26B A4B it 25.2B	Tight fit	UD-IQ4_XS	8K	42–58
bottlecapai ThinkingCap gemma 4 26B A4B it 25.2B	Tight fit	IQ4_XS	8K	41–56
GLM 4.7 Flash 29.9B	Tight fit	UD-Q3_K_XL	8K	52–71
Qwen3.5 35B A3B 34.7B	Tight fit	UD-IQ3_S	8K	131–179
Qwen3.6 35B A3B 34.7B	Tight fit	UD-IQ3_S	8K	130–178
Ministral 3 14B Instruct 2512 13.5B	Tight fit	Q8_0	4K	12–16
Ministral 3 14B Reasoning 2512 13.5B	Tight fit	Q8_0	4K	12–16
Qwen3 Coder 30B A3B Instruct 30.5B	Tight fit	UD-Q3_K_XL	8K	81–111
Qwen3 30B A3B Instruct 2507 30.5B	Tight fit	UD-Q3_K_XL	8K	81–111
Qwen3 VL 30B A3B Instruct 30.5B	Tight fit	UD-Q3_K_XL	8K	81–111
Qwen3 30B A3B Thinking 2507 30.5B	Tight fit	UD-Q3_K_XL	8K	81–111
gemma 4 12b it 11.9B	Tight fit	UD-Q8_K_XL	8K	10–14
Tiel Coder 35B A3B 34.7B	Runs	UD-IQ3_XXS	8K	134–183
Kwaipilot KAT Coder V2.5 Dev 34.7B	Tight fit	IQ3_XXS	8K	61–84
Qwen3.6 14B A3B FableVibes 13.8B	Tight fit	Q8_0	8K	52–71
GLM 4.6V Flash 9.4B	Runs	UD-Q8_K_XL	8K	14–19
GLM Z1 9B 0414 9.4B	Runs	UD-Q8_K_XL	8K	14–19

105 more models in the catalog do not fit this default profile, or ranked below the first 40 that do.

Example: Tight fit fully on your NVIDIA T4 at UD-Q4_K_XL, 4K context, about 12–16 tok/s. Weights 13.5 GB plus 0.63 GB of KV cache — 14.6 GB of 15.2 GB usable. Weights shown in the verdict use GiB (the same 1024³ counting as GPU memory).