Hardware

GeForce RTX 3070 Ti

GeForce RTX 3070 Ti has 8 GB GDDR6X and 608 GB/s. See which open models fit, at which quant, and roughly how fast. Figures below use the default memory (8 GB), 32 GB system RAM, and DDR5-5600 for anything that spills. Updated 2026-10-24.

Memory8 GB GDDR6X

Bandwidth608 GB/s

Kindgpu

Year2021

NVIDIA GeForce RTX 30-series specifications. Source.

RSS for new models on this device

Models that fit

Model	Verdict	Quant	Context	tok/s
GLM 4.6V Flash 9.4B	Tight fit	Q5_K_S	8K	48–65
GLM Z1 9B 0414 9.4B	Tight fit	Q5_K_S	8K	48–65
GLM 4 9B 0414 9.4B	Tight fit	Q5_K_S	8K	48–65
Qwen3.8 9B Distill 9.2B	Tight fit	Q5_K_M	8K	48–66
Qwythos 9B Claude Mythos 5 1M 9.0B	Tight fit	Q5_K_M	8K	50–68
Qwythos 9B v2 9.0B	Tight fit	Q5_K_M	8K	49–67
Qwen3.5 9B 9.0B	Tight fit	UD-Q5_K_XL	8K	48–65
MiMo V2.6 Distill Qwen 9B 9.0B	Tight fit	Q5_K_M	4K	48–65
granite 4.2 8b 8.8B	Tight fit	Q5_K_M	4K	48–66
gemma 4 12b it 11.9B	Tight fit	IQ4_XS	4K	42–57
Ministral 3 8B Instruct 2512 8.5B	Tight fit	UD-Q5_K_XL	4K	50–69
Ministral 3 8B Reasoning 2512 8.5B	Tight fit	UD-Q5_K_XL	4K	50–69
Qwen3.8 27B 27.3B	Tight fit	UD-IQ1_M	4K	48–65
gemma 3 27b it 27.0B	Tight fit	UD-IQ1_S	1K	48–65
Qwen3 VL 8B Instruct 8.2B	Tight fit	UD-Q5_K_XL	8K	47–64
DeepSeek R1 0528 Qwen3 8B 8.2B	Tight fit	UD-Q5_K_XL	8K	47–64
Qwen3 8B 8.2B	Tight fit	UD-Q5_K_XL	8K	47–64
Qwen3 VL 8B Thinking 8.2B	Tight fit	UD-Q5_K_XL	8K	47–64
medgemma 27b text it 27.0B	Tight fit	UD-IQ1_S	2K	44–60
DeepSeek R1 Distill Llama 8B 8.0B	Tight fit	UD-Q5_K_XL	8K	49–67
Llama 3.1 8B Instruct 8.0B	Tight fit	UD-Q5_K_XL	8K	49–67
Apertus 8B Instruct 2509 8.1B	Tight fit	UD-Q5_K_XL	8K	48–66
Qwen2.5 7B Instruct 7.6B	Runs	Q5_K_M	8K	57–77
Qwen2.5 VL 7B Instruct 7.6B	Tight fit	Q6_K	8K	50–68
DeepSeek R1 Distill Qwen 7B 7.6B	Tight fit	Q6_K	8K	50–68
Qwen2.5 Coder 7B Instruct 128K 7.6B	Tight fit	Q6_K	8K	50–68
NVIDIA Nemotron 3.5 Bolt E4B 7.5B	Runs	UD-Q5_K_XL	8K	51–70
gemma 4 E4B it 7.5B	Tight fit	UD-Q5_K_XL	8K	45–62
Devstral Small 2 24B Instruct 2512 23.6B	Tight fit	UD-IQ1_M	4K	50–68
Mistral Small 3.2 24B Instruct 2506 23.6B	Tight fit	UD-IQ1_M	4K	50–68
Magistral Small 2509 23.6B	Tight fit	UD-IQ2_XXS	2K	47–64
Qwen2.5 14B Instruct 14.8B	Tight fit	Q3_K_S	2K	47–65
gemma 4 E4B it qat 7.5B	Runs	UD-Q4_K_XL	8K	68–93
Olmo 3 7B Instruct 7.3B	Tight fit	UD-Q4_K_XL	4K	50–68
Qwen3 14B 14.8B	Tight fit	UD-IQ3_XXS	4K	50–68
Ministral 3 14B Instruct 2512 13.5B	Tight fit	Q3_K_S	4K	50–68
Ministral 3 14B Reasoning 2512 13.5B	Tight fit	Q3_K_S	4K	50–68
LFM2.5 8B A1B 8.5B	Tight fit	UD-Q5_K_XL	8K	289–394
Qwen3.6 14B A3B FableVibes 13.8B	Tight fit	Q3_K_M	8K	204–278
NousResearch Hermes 4 14B 14.8B	Tight fit	IQ3_XXS	4K	51–69

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 GeForce RTX 3070 Ti at Q5_K_S, 8K context, about 48–65 tok/s. Weights 6.24 GB plus 0.31 GB of KV cache — 6.91 GB of 7.13 GB usable. Weights shown in the verdict use GiB (the same 1024³ counting as GPU memory).