AI BriefingKakao Releases Open-Source 'Kanana 1.5' for Agentic AI with Major Math and Coding Enhancements
Key point
Kakao has open-sourced the 'Kanana 1.5' 8B and 2.1B models under the Apache 2.0 license, enhancing math and coding performance and long-context understanding to support Agentic AI.
Details
Kakao released version 1.5 of the Kanana LLM to address the era of Agentic AI. This update focuses particularly on advancing math and coding capabilities, with the Kanana-1.5-8B Instruct model recording 76.83 on HumanEval+ and 67.54 on MATH, showing significant performance improvements over previous versions. Additionally, context length was extended to 128K through Position Interpolation and RoPE variants to improve long-document processing capabilities. The released models are available in 8B and 2.1B parameter sizes under the Apache 2.0 license, which permits commercial use. Based on this, Kakao plans to focus on developing a more advanced Kanana 2 in the future.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Kakao Ends Beta of Standalone AI App 'Kanana' to Shift to Integrated Services Like KakaoTalk