Seeking User Experiences and Feedback

#2
by BuiDoan - opened

Hi everyone,

I’ve come across a language model (LLM) that seems promising, but there’s very limited public information available about it. Has anyone had hands-on experience with this model?

Could you share your thoughts on:

Performance and reliability in real-world tasks
Inference speed and resource requirements
Strengths and limitations compared to more well-known models
Any insights or personal experiences would be really helpful! Thanks in advance for your time and feedback.

I am currently trying to deploy this variant, I have had Apriel-1.5-15b-thinker running via vLLM and it is comparable to GPT-4o or better depending on configuration, system prompt and tool calling. I personally wrote a custom 3 pass router for tool calls that has multi-faceted prompts throughout. The model itself is surprisingly fast, intelligent and very nuanced until you tweak it.

I can't get this version to run via vLLM.

Sign up or log in to comment