"Isaacus, an Australian foundational legal AI startup, has launched Kanon 2 Embedder, a state-of-the-art legal embedding LLM, and unveiled the Massive Legal Embedding Benchmark (MLEB), an open-source benchmark for evaluating legal information retrieval performance across six jurisdictions (the US, UK, EU, Australia, Singapore, and Ireland) and five domains (cases, statutes, regulations, contracts, and academia).
Kanon 2 Embedder ranks first on MLEB as of 23 October 2025, delivering 9% higher accuracy than OpenAI Text Embedding 3 Large and 6% higher accuracy than Google Gemini Embedding while running >30% faster than both LLMs. Kanon 2 Embedder leads a field of 20 LLMs, including Qwen3 Embedding 8B, IBM Granite Embedding R2, and Microsoft E5 Large Instruct."
"We note, however, that there is, unfortunately, a serious risk that Voyage’s models were trained on some of the evaluation sets in MLEB, particularly SCALR and Consumer Contracts QA, which are both also part of MTEB, due to the fact that Voyage trains on their customers’ private data by default (which would invariably include benchmarks). This is also a risk for Cohere and Jina models."
Kanon 2 Embedder ranks first on MLEB as of 23 October 2025, delivering 9% higher accuracy than OpenAI Text Embedding 3 Large and 6% higher accuracy than Google Gemini Embedding while running >30% faster than both LLMs. Kanon 2 Embedder leads a field of 20 LLMs, including Qwen3 Embedding 8B, IBM Granite Embedding R2, and Microsoft E5 Large Instruct."