DefiledAI Research

MODEL DATABASE

Open-weight models catalogued by family, with quantization options, context windows, and minimum VRAM requirements for local inference.

Llamaby Meta

ACTIVE

Model	Params	Context	Min VRAM	License	Quants Available
Llama 3.1 8B	8B	128K	6GB	Llama 3	Q4_K_MQ5_K_MQ8_0F16
Llama 3.1 70B	70B	128K	40GB	Llama 3	Q2_KQ4_K_MQ5_K_MIQ3_M
Llama 3.1 405B	405B	128K	240GB	Llama 3	Q2_KIQ1_M

Qwenby Alibaba

ACTIVE

Model	Params	Context	Min VRAM	License	Quants Available
Qwen 3 7B	7B	32K	6GB	Apache 2.0	Q4_K_MQ5_K_MQ8_0
Qwen 3 14B	14B	32K	10GB	Apache 2.0	Q4_K_MQ5_K_M
Qwen 3 72B	72B	32K	40GB	Apache 2.0	Q2_KQ4_K_MQ5_K_M

DeepSeekby DeepSeek

ACTIVE

Model	Params	Context	Min VRAM	License	Quants Available
DeepSeek R1 7B	7B	32K	6GB	MIT	Q4_K_MQ8_0
DeepSeek R1 70B	70B	32K	40GB	MIT	Q4_K_MIQ3_M
DeepSeek V3	671B MoE	128K	Multi-GPU	MIT	Q2_KIQ1_M

Mistralby Mistral AI

ACTIVE

Model	Params	Context	Min VRAM	License	Quants Available
Mistral 7B v0.3	7B	32K	6GB	Apache 2.0	Q4_K_MQ5_K_MQ8_0F16
Mixtral 8x7B	56B MoE	32K	24GB	Apache 2.0	Q4_K_MQ5_K_M
Mixtral 8x22B	141B MoE	64K	48GB	Apache 2.0	Q2_KQ4_K_M

Gemmaby Google

ACTIVE

Model	Params	Context	Min VRAM	License	Quants Available
Gemma 2 2B	2B	8K	2GB	Gemma	Q4_K_MQ8_0F16
Gemma 2 9B	9B	8K	6GB	Gemma	Q4_K_MQ5_K_MQ8_0
Gemma 2 27B	27B	8K	16GB	Gemma	Q4_K_MQ5_K_M

Phiby Microsoft

ACTIVE

Model	Params	Context	Min VRAM	License	Quants Available
Phi-3 Mini	3.8B	128K	3GB	MIT	Q4_K_MQ8_0F16
Phi-3 Medium	14B	128K	10GB	MIT	Q4_K_MQ5_K_M