KL3M
Click to visit website
About
KL3M is a family of clean large language models designed to avoid IP, toxicity, and copyright issues. It's trained on high-quality, transparently sourced data, explicitly excluding content with legal or ethical concerns, such as copyrighted material, scraped websites violating ToS, or synthetic data from other LLMs. KL3M is the first language model to receive the Fairly Trained L Certification, ensuring respect for content creators' rights. Its models, including `kl3m-170m` and `kl3m-1.7b`, demonstrate best-in-class perplexity on business content like legal and financial material, outperforming models with 10x parameters in some cases. They also exhibit the lowest rates of "bad" words and toxicity. KL3M models are used for tasks like drafting invoices, contracts, SEC filings, and patents. Users can continue training, fine-tune for conversational AI or specific tasks, or license its 2.5T+ underlying training data. Future models will support more languages beyond English.
Platform
Task
Features
• supports fine-tuning and continued pretraining
• high efficiency for legal and financial content
• low toxicity and bias in models
• fairly trained l certification
• absence of toxic content
• exclusion of llm synthetic data
• no copyright infringement or tos violations
• clean provenance and data sourcing
FAQs
What kind of hardware do I need to run KL3M?
The first KL3M models have been designed with accessible use as a priority. kl3m-170 runs quickly on a MacBook Air M1, and kl3m-1.7b runs well on a $300 consumer GPU.
What architectures are your models?
Smaller KL3M models are trained using the GPT-NeoX architecture. Larger KL3M models are trained using the Mixtral Mixture-of-Experts architecture (trained from scratch).
How can I run KL3M?
KL3M is distributed as standard PyTorch model weights. KL3M architectures are supported for both HuggingFace transformers and vllm for inference.
Which languages are supported?
`kl3m-170m` and `kl3m-1.7b` have both been trained on a predominantly English-language content. Larger models include content in English, Spanish (es-ES and es-MX), French, and German.
Do you provide an API?
Not yet. Our focus has been on enabling the use of small, local LLMs for information security and accessibility purposes, but we are evaluating the possibility of providing an API in the future.
Is it easy to fine-tune KL3M?
We have had excellent results fine-tuning KL3M on a number of use cases, including drafting, summarization, and classification. You can fine-tune kl3m-170 and kl3m-1.7b on consumer hardware.
How many tokens do you have?
We have collected over 2.5 trillion tokens of training data, and we are constantly adding more. Our training data is a mix of public domain and explicitly licensed content.
How many tokens have your models seen?
`kl3m-170m` and `kl3m-1.7b` have been trained on approximately 350B tokens of primarily English-language content. Larger models are being trained on between 500B to 1T tokens of content in English, Spanish, French, and German.
Do you have a conversational chat model?
Not yet. While our pretraining data does include a number of conversational sources, we have not yet trained a model that is designed for standard conversational rounds. Stay tuned.
Do you have a general instruction-aligned model?
Our base models already support a number of tasks like extractive/abstractive summarization or conversion, but we have not trained an open-ended model. Our first instruct model supports legal drafting and revision, and we'd love to hear what other tasks you'd like supported.
How do you pronounce KL3M?
As the 🍊 suggests, KL3M is pronounced like "Clem" or "Klem."
Why is it named KL3M?
KL3M was originally short for the Kelvin Legal Large Language Model, KLLLM. Because we're nerds, we shortened all those Ls to L cubed or L3, then shortened K-L3-M to KL3M.
Job Opportunities
There are currently no job postings for this AI tool.
Ratings & Reviews
No ratings available yet. Be the first to rate this tool!
Alternatives
Spellbook
Spellbook is an AI-powered tool for commercial lawyers that helps draft and review contracts faster within Word. It provides features for redlining, drafting, answering complex questions, and benchmarking contracts.
View Details
Leagle.AI
Leagle.AI is your digital assistant for drafting legal documents, helping you create tailored legal documents in real-time based on your specific needs.
View Details
AI.Law
AI.Law is an advanced AI legal platform that revolutionizes litigation by drafting legal documents, reports, and discovery with unprecedented speed and accuracy.
View Details
DocDraft
DocDraft is an AI-powered legal tool that automates the creation of contracts, agreements, and legal documents. It offers expert review and instant legal support, making legal matters easier for individuals and businesses.
View Details
SAVVY.AI
SAVVY.AI is the first AI-powered application in Hong Kong that drafts legally binding legal documents free of charge, prepared by legal professionals for online signing.
View DetailsFeatured Tools
GirlfriendGPT
NSFW AI chat platform with customizable characters, AI image generation, and voice chat. Explore roleplay and intimate interactions with AI companions.
View Details
Animate My Pic
Animate My Pic is an AI photo to video tool that leverages advanced AI to effortlessly animate your pictures, offering image-to-video, text-to-video, and 30+ effects.
View Details
Nano Banana AI
Nano Banana AI is a powerful AI image editor for quick, precise editing, adjustments, and optimization of images, leveraging advanced image-to-image AI models.
View Details
Nano Banana
Nano Banana is Google's state-of-the-art AI image generator powered by Gemini 2.5 Flash Image, offering character consistency and natural language image transformation.
View Details
alivemoment
alivemoment is an AI tool that transforms cherished photos into living stories, allowing users to relive precious moments with gentle, lifelike motion.
View Details