Best Open-Source Model for Agentic Apps with CrewAI

:magnifying_glass_tilted_left: Looking for the Best Open-Source Model for Agentic Apps with CrewAI

Hey Hugging Face community! :waving_hand:

I’m currently building an Agentic App using CrewAI, and I’m looking for open-source LLMs that are well-suited for agent-based reasoning and tool use.

I’m specifically looking for:

:brain: Models optimized for agentic behavior, tool use, or multi-agent collaboration

:rocket: Performance close to GPT-4o (or as close as possible in open-source)

:light_bulb: Models with Instruct / Chat tuning

:feather: Preferably small to medium models (e.g. 7B–13B) that I can run locally or with reasonable hardware

:globe_showing_europe_africa: Open-source and easy to integrate with tools like llama.cpp, ollama, or vLLM

If you’ve tested any models that work particularly well in an agentic setup or have had great results in CrewAI-based workflows, I’d love your suggestions! :folded_hands:

Thanks in advance for your help, and I’ll be happy to share my results with the community once I benchmark a few!

1 Like

The Mistral Small 2503 seems suitable, but it might be a bit too large (24B). If you’re looking for a 7B/8B class, I would personally recommend the Qwen 2.5 series.

https://huggingface.co/spaces/open-llm-leaderboard