AI tools for Open Source Models
Browse 14 catalog entries for open source models tasks. Listings are displayed alphabetically and are not quality, popularity, or editorial rankings.
Catalog entries
Command R+
Cohere's retrieval-grounded open-weights giant optimized for multilingual secure RAG.
DeepSeek
Leaderboard-disrupting reasoning model series with fully open R1 weights files.
DeepSeek-V3 Open
A massively routed, cost-disruptive Mixture-of-Experts model now superseded by DeepSeek V4.
Gemma 2 27B
Google's open-weights model utilizing custom distillation, rivaling double parameter models.
Llama
Meta's highly optimized open-weights models enabling completely private local hosting.
Llama 3 8B
Meta's lightweight, high-performance open model, optimized for local laptop installations.
Llama 3.1 405B
Meta's flagship open-weights giant from 2024, now superseded by Llama 4 for new deployments.
Mistral
European open-weights architectures noted for extreme Mixtral MoE compute efficiency.
Mistral Nemo 12B
Mistral's multilingual powerhouse developed in collaboration with NVIDIA, featuring excellent context.
Mixtral 8x22B
Mistral's Mixture-of-Experts (MoE) open-weights powerhouse with exceptional throughput.
Phi-3.5 Mini
Microsoft's lightweight 3.8B model outperforming giants in logical math reasoning tests.
Qwen3-Coder-Next 80B
Alibaba's dedicated open-weights developer model, rivaling closed systems in code formatting.
StarCoder 2
The BigCode research model trained strictly on licensed, public-domain codebases.
Whisper Large V3
OpenAI-trained vocal transcription engine translating multi-language spoken files.
Our Methodology
This beta presents documented catalog details without public scores or rankings. Check each listing's official source and verification date before relying on it.
Read Evaluation Standards