Can the GTX 1660 / 1660 Super (6 GB) run local LLMs? Models, VRAM and speed

Best picks for the GTX 1660 / 1660 Super

Best for chat and writingQwen3.5 4BQ6_K · 5.1 GB · ~57 tok/s Best for codingQwen3 4BQ6_K · 4.7 GB · ~67 tok/s Best for reasoning and mathsPhi-4 Mini 3.8BQ8_0 · 5.2 GB · ~54 tok/s Best for image understandingQwen3.5 4BQ6_K · 5.1 GB · ~57 tok/s

Every model on the GTX 1660 / 1660 Super

Results at 8K context, picking the highest quality quantization that fits. "Offload" means part of the model runs from system RAM (we assume 32 GB), which works but is much slower.

Model	Status	Quant	Memory	Speed
Qwen3.5 4B	Runs well	Q6_K	5.1 GB	~57 tok/s
Gemma 3 4B	Runs well	Q5_K_M	5.3 GB	~72 tok/s
Qwen3 4B	Runs well	Q6_K	4.7 GB	~67 tok/s
Phi-4 Mini 3.8B	Runs well	Q8_0	5.2 GB	~54 tok/s
Llama 3.2 3B	Runs well	Q8_0	4.6 GB	~64 tok/s
Llama 3.2 1B	Runs well	Q8_0	1.6 GB	~170 tok/s
Gemma 3 1B	Runs well	Q8_0	1.4 GB	~210 tok/s
Qwen3 0.6B	Runs well	Q8_0	0.9 GB	~340 tok/s
Mistral 7B v0.3	Offload	Q4_K_M	5.7 GB	~7.7 tok/s
Qwen2.5-Coder 7B	Offload	Q4_K_M	5.8 GB	~7.3 tok/s
Llama 3.1 8B	Offload	Q4_K_M	6.2 GB	~7.0 tok/s
DeepSeek-R1 Distill 8B	Offload	Q4_K_M	6.2 GB	~7.0 tok/s
Qwen3 8B	Offload	Q4_K_M	6.4 GB	~6.8 tok/s
Qwen3.5 9B	Offload	Q4_K_M	7.0 GB	~6.0 tok/s
Mistral Nemo 12B	Offload	Q4_K_M	9.2 GB	~4.7 tok/s
Gemma 4 12B	Offload	Q4_K_M	11 GB	~4.8 tok/s
Phi-4 14B	Offload	Q4_K_M	11 GB	~3.9 tok/s
Gemma 3 12B	Offload	Q4_K_M	11 GB	~4.7 tok/s
Qwen3 14B	Offload	Q4_K_M	11 GB	~3.9 tok/s
DeepSeek-R1 Distill 14B	Offload	Q4_K_M	11 GB	~3.9 tok/s
gpt-oss 20B	Offload	Q4_K_M	14 GB	~11 tok/s
Codestral 22B	Offload	Q4_K_M	16 GB	~2.6 tok/s
Devstral Small 24B	Offload	Q4_K_M	17 GB	~2.4 tok/s
Mistral Small 3 24B	Offload	Q4_K_M	18 GB	~2.4 tok/s
Gemma 4 26B-A4B (MoE)	Offload	Q4_K_M	19 GB	~10 tok/s
Qwen3.6 27B	Offload	Q4_K_M	19 GB	~2.1 tok/s
Qwen3 30B-A3B (MoE)	Offload	Q4_K_M	21 GB	~11 tok/s
Qwen3-Coder 30B-A3B	Offload	Q4_K_M	21 GB	~11 tok/s
Gemma 4 31B	Offload	Q3_K_M	21 GB	~2.3 tok/s
Gemma 3 27B	Offload	Q4_K_M	22 GB	~2.1 tok/s
Qwen3 32B	Offload	Q4_K_M	23 GB	~1.8 tok/s
DeepSeek-R1 Distill 32B	Offload	Q4_K_M	23 GB	~1.8 tok/s
Qwen2.5-Coder 32B	Offload	Q4_K_M	23 GB	~1.8 tok/s
Qwen3.6 35B-A3B (MoE)	Offload	Q4_K_M	23 GB	~12 tok/s
Llama 3.3 70B	Won’t fit	—	48 GB	—
DeepSeek-R1 Distill 70B	Won’t fit	—	48 GB	—
Llama 4 Scout 109B-A17B	Won’t fit	—	73 GB	—
gpt-oss 120B	Won’t fit	—	77 GB	—
Qwen3 235B-A22B (MoE)	Won’t fit	—	155 GB	—

What this means in practice

The GTX 1660 / 1660 Super is comfortable with models up to about 4B parameters. gpt-oss 20B is the most useful step up beyond that: it needs about 14 GB, so part of it runs from system RAM, but as a mixture-of-experts model it stays reasonably quick at around 11 tokens per second. Models from Llama 3.3 70B upward need 48 GB or more and are out of reach. NVIDIA cards have the broadest support: Ollama, LM Studio and llama.cpp all use CUDA out of the box.

The next memory tier up is 8 GB: Arc A750, RTX 5060 Ti, RTX 5060. Other 6 GB cards: RTX 4050 Laptop, RTX 3060 Laptop.

Compare the GTX 1660 / 1660 Super

Compare with any cardPick the other one yourself

Questions

What is the biggest model the GTX 1660 / 1660 Super can run?

At 8K context the largest model that fits entirely in its 6 GB of VRAM is Qwen3.5 4B at Q6_K, at roughly 57 tokens per second.

Can the GTX 1660 / 1660 Super run 70B models like Llama 3.3 70B?

No. It needs about 48 GB at Q4, which is far beyond 6 GB even with system RAM offload.

What is the best model for coding on the GTX 1660 / 1660 Super?

Qwen3 4B at Q6_K fits fully and runs at about 67 tokens per second. Pair it with an editor extension like Continue pointed at Ollama.

Estimates based on memory size and bandwidth; real speed varies with drivers, software and context length. See how we calculate VRAM.

Can the GTX 1660 / 1660 Super run local LLMs?

Check the GTX 1660 / 1660 Super live

Models 0 of 0 run