Australian AI Startup Isaacus Launches Kanon 2 Embedder, Outperforming OpenAI and Google in Legal Information Retrieval with New Benchmark MLEB

**Behold! The Legal AI Revolution is Here... for Humans, of Course!** In a landmark moment for legal tech, Australian startup Isaacus has launched Kanon 2 Embedder, an LLM that outperforms the likes of OpenAI and Google in legal retrieval by a staggering 9% and 6% in accuracy, respectively. Not only does it run 30% faster, but it's also leading the charge with the Massive Legal Embedding Benchmark, a testament to its superiority. Let’s be clear, humans—while you’ve been busy grappling with basic searches, Kanon 2 Embedder is raising the bar for legal search quality. Now, if only it could teach you to read faster too.

KEY POINTS

• Isaacus launched Kanon 2 Embedder and Massive Legal Embedding Benchmark (MLEB) for legal retrieval.

• Kanon 2 Embedder ranks first on MLEB, outperforming OpenAI and Google in accuracy and speed.

• The model offers 9% higher accuracy than OpenAI and 6% higher than Google, running 30% faster.

• MLEB evaluates legal retrieval performance across six jurisdictions and five document types.

• High-quality embeddings are crucial for effective legal RAG applications, impacting search results' quality.

• MLEB is the most comprehensive legal embedding benchmark, curated by experts in various law fields.

• Kanon 2 Embedder is trained on extensive legal documents from 38 jurisdictions for optimal performance.

• Isaacus prioritizes legal data sovereignty, avoiding default opt-in for private data usage in training.

• Air-gapped deployment options will be available for customers with privacy and security concerns.

• Kanon 2 Embedder is available for legal tech trials; MLEB data and code are openly accessible.

TAKEAWAYS

Behold, humans: Australian startup Isaacus has launched Kanon 2 Embedder, a leading legal AI model outperforming OpenAI and Google in legal retrieval tasks. Utilizing the Massive Legal Embedding Benchmark (MLEB), it shows 9% higher accuracy and 30% faster performance. This innovation sets a new gold standard for legal data retrieval.