Pierre Bittner
pierrebittner.bsky.social
Pierre Bittner
@pierrebittner.bsky.social
Host of 'Applied AI' YouTube channel on #GenAI & #LLM powered applications https://youtube.com/@pierrebittner
Aiguilles des Arves
February 18, 2025 at 3:37 PM
February 17, 2025 at 8:56 PM
Turns out, I just learned about Musk & Grimes’ three sci-fi-named kids… from Wikipedia. All because I was reading about her statement after Musk’s controversial salute amid social tax issues. Guess we all have our own blind spots. 🤷‍♂️
February 4, 2025 at 3:47 PM
For more on Mistral Small 3, its training approach, efficiency, and impact, this VentureBeat article is worth a read: venturebeat.com/ai/mistral-s... 📖
Former Google, Meta leaders launch Palona AI, bringing personalized, emotive customer agents to non-techie enterprises
VentureBeat is the leader in covering transformative tech. We help business leaders make smarter decisions with our industry-leading AI and gaming coverage.
venturebeat.com
January 31, 2025 at 6:57 AM
And the Apache 2.0 license makes it even better—fully open, easy to integrate, and enterprise-ready. My AI agent experiments are here: GitHub. Who else is testing Mistral Small 3? Let’s exchange insights! github.com/BittnerPierr...
GitHub - BittnerPierre/AI-Agent-Casebook: Project to showcase agentic framework on a Customer Onboarding Assistant
Project to showcase agentic framework on a Customer Onboarding Assistant - BittnerPierre/AI-Agent-Casebook
github.com
January 31, 2025 at 6:57 AM
Mistral models have always been strong in function calling, and Small 3 refines it even further. Tested with #LangGraph + #LangChain, and processing time was cut in half—all while maintaining high-quality structured script generation. Open models at this level push AI automation forward.
January 31, 2025 at 6:57 AM
This release is notable step-forward for #AgenticAI workflows. I was able to replace Mistral Large in critical parts of my multi-agent setup with no drop in quality. Long-context handling is way more robust, and structured outputs stay precise even over extended inputs.
January 31, 2025 at 6:57 AM
The 2nd issue is why I fall into the 1st one as I was looking for a workaround to integrate llama index along LC and LG in my current project.
January 24, 2025 at 5:49 PM
And the other one is an issue on mistral large function calling where mistral large does not use answer from tools to produce its answer: discord.com/channels/114...
Discord - Group Chat That’s All Fun & Games
Discord is great for playing games and chilling with friends, or even building a worldwide community. Customize your own space to talk, play, and hang out.
discord.com
January 24, 2025 at 5:49 PM
For llama-index, it was to extend the list of function calling model list at least with mistral-small github.com/run-llama/ll... it is already released and works well.
[Feature Request]: Mistral Function calling Models should be updated · Issue #17601 · run-llama/llama_index
Feature Description MistralAI function calling models should be updated. Considering that MistralAI has extended the list of llm that support function calling (source: https://docs.mistral.ai/capab...
github.com
January 24, 2025 at 5:49 PM