Building with Mistral
summary
Mistral lays out exact models, licenses and demos: pixol 12B is open-weight Apache 2.0, multimodal with a 128k context window and image-to-JSON demos, one extracting a receipt total of 68; Mr Large 123B is the flagship for reasoning and code. The platform offers function calling, a LoRA-style fine-tuning pipeline and hosted fine-tune API, an agents beta and an open-source finetune repo, and claims models fit on a single ed100 node for convenient on-prem or cloud deployment.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 1:53Mistral model offerings
- 6:52Mistral Large 2
- 10:06Pixtral multimodal model
- 18:43Customizing Mistral models
- 20:53Fine-tuning for domain tasks
- 23:26Autonomous agents with Mistral
- 29:50Community resources
- 31:18Q&A
related talks
13mKeynote: Building with Mistral
34mA Journey into Open Source Models with HuggingFace & Mistral
29mLeveraging open-source LLMs for production
19m2025 in LLMs so far, illustrated by Pelicans on Bicycles
24mTeaching Mistral to Reason: Post-Training with PyTorch and NVIDIA
25mSelecting the Best LLMs with Model Cockpit
49mOpen Models: The Secret Weapon for Next-Generation Software
31mDeploying Large Language Models with Ease