Trip Venturella Releases Mr. Chatterbox Victorian LLM
Trip Venturella released Mr. Chatterbox on March 30, 2026, a 340 million-parameter language model trained solely on a British Library corpus of 28,035 Victorian texts (1837–1899), totaling about 2.93 billion tokens. The 2.05GB model is available via a Hugging Face Spaces demo and an llm plugin, but the author found its responses limited, illustrating public-domain-only models need much more data to reach conversational quality.
Key Points
- 1Trains a 340M-parameter LLM on 28,035 British Library texts totaling 2.93B tokens.
- 2Highlights data limitation: Chinchilla scaling implies ~7B tokens needed for comparable performance.
- 3Suggests practitioners require larger public-domain corpora or hybrid data to reach usable chat quality.
Scoring Rationale
This is a timely, hands-on release showing a working public-domain-only LLM and runnable plugin, raising practical interest. Score reflects moderate novelty and strong actionability for practitioners, limited scope and single-source credibility, and the clear data不足 relative to Chinchilla scaling.
Sources
Public references used for this report.
Practice with real Logistics & Shipping data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Logistics & Shipping problems
