ChatLlama 3 Community
Llama 3 70B
Meta · Llama
Meta's flagship open weights model. Strong reasoning, coding, and instruction following.
VRAM48GBQ416GB (Q4)
Details →Open Source AI Models
Filterable directory of open source models: hardware requirements, licenses, and quick-start guides.
AllChatCodeEdgeAudio
ChatApache 2.0
Mistral 7B
Mistral AI · Mistral
Punches well above its weight class. Excellent for local inference on consumer GPUs.
VRAM8GBQ44GB (Q4)
Details →ChatQwen License
Qwen2 72B
Alibaba · Qwen
Top-tier open weights model with strong multilingual and coding capabilities.
VRAM48GBQ424GB (Q4)
Details →CodeDeepSeek License
DeepSeek Coder V2
DeepSeek · DeepSeek
Best-in-class open source coding model. Rivals GPT-4o on coding benchmarks.
VRAM80GBQ432GB (Q4)
Details →EdgeMIT
Phi-3 Mini
Microsoft · Phi
3.8B parameter model that outperforms models 3x its size. Runs on CPU.
VRAM4GBQ42GB (Q4)
Details →ChatGemma Terms
Gemma 2 27B
Google · Gemma
Google's open weights model family. Strong performance at the 27B scale.
VRAM24GBQ412GB (Q4)
Details →CodeLlama 2 Community
Code Llama 34B
Meta · Llama
Meta's code-specialized model. Strong at completion, infilling, and instruction following.
VRAM24GBQ412GB (Q4)
Details →ChatApache 2.0
Falcon 40B
TII · Falcon
UAE Technology Innovation Institute open model. Strong multilingual capabilities.
VRAM32GBQ416GB (Q4)
Details →ChatStability AI NC
StableLM 2 12B
Stability AI · StableLM
Efficient chat model with good balance of performance and resource requirements.
VRAM12GBQ46GB (Q4)
Details →AudioMIT
Whisper Large V3
OpenAI · Whisper
State of the art open source speech recognition. Multilingual transcription and translation.
VRAM10GB
Details →