Hugging Face is officially retiring the ‘NLP Course’ branding that has been a cornerstone of open-source AI education for three years. The revamped resource, now called ‘The LLM Course,’ reflects a curriculum that has outgrown its original classification and entity-recognition roots. The team found that new chapters on fine-tuning large language models and building reasoning models like DeepSeek R1 simply didn’t fit under the old NLP banner.
The shift isn’t just cosmetic. It follows the experimental success of the smol-course and a massive agents course that drew 100,000 registrations, proving there’s a huge appetite for interactive, modern AI education. The company is committing to a more framework-agnostic approach going forward. While the Hugging Face transformers library remains the reference implementation, new chapters on fine-tuning and inference are built in collaboration with external libraries and tools that complement the ecosystem. The goal is to reflect the actual tools developers are mixing and matching in production, not just the Hugging Face stack.
That doesn’t mean the classic material is getting thrown out. Hugging Face is keeping the existing chapters on tasks like named-entity recognition and retrieval, arguing that not every problem requires a massive model. Students still benefit from simpler, interpretable tasks that run locally. However, even these older sections are getting a modernization pass, with planned updates to include Sentence Transformers, Zero Shot classification techniques, and the ModernBERT architecture.
Looking ahead, the course will continue expanding with interactive exercises and live sessions reserved for topics where synchronous engagement actually adds value. The core focus remains on durable written material and coding exercises. Hugging Face is opening the doors for community contributions, inviting authors, maintainers, and companies to propose new units or host live sessions through the organization’s Hub. It’s a pragmatic update that acknowledges the market has moved decisively toward LLMs, without pretending that classic NLP has suddenly become irrelevant.
