Snowflake launches Arctic Embed 2.0, an open-weights multilingual embedding model optimised for APAC enterprise search — outperforming OpenAI text-embedding-3-large on Chinese, Japanese, Korean, and Bahasa Indonesia retrieval benchmarks at 40% lower inference cost.
Snowflake has launched Arctic Embed 2.0, the latest version of its open-weights embedding model series, specifically optimised for multilingual APAC enterprise search workloads including Chinese, Japanese, Korean, Bahasa Indonesia, and Bahasa Malaysia retrieval tasks.
Arctic Embed 2.0 achieves state-of-the-art results on the BEIR multilingual benchmark for four APAC languages evaluated: it outperforms OpenAI's text-embedding-3-large on Chinese document retrieval (CMTEB benchmark, +3.2 points nDCG@10), Japanese legal document search (+2.8 points), Korean financial report retrieval (+1.9 points), and Bahasa Indonesia product catalogue search (+4.1 points). The model runs at 40% lower inference cost than text-embedding-3-large when self-hosted on APAC cloud infrastructure, with quantised INT8 variants available for further cost reduction on high-throughput APAC enterprise search workloads.
For APAC ML engineering and data teams building enterprise search, RAG systems, and semantic matching applications, Arctic Embed 2.0 is immediately relevant for three use cases: document search over APAC regulatory filings and contracts in mixed Chinese-English content, product catalogue search for Southeast Asian e-commerce with Bahasa Indonesia and Bahasa Malaysia queries, and multilingual customer support ticket routing where APAC customer messages arrive in multiple languages across a single support queue. Snowflake is making Arctic Embed 2.0 available via Snowflake Cortex AI — usable directly in Snowpark Python without exporting data to external ML serving infrastructure.
Beyond this story
Cross-reference our practice depth.
News pieces sit on top of working capability. Browse the service pillars, industry verticals, and Asian markets where AIMenta turns these stories into engagements.
Other service pillars
By industry
Other Asian markets
Related stories
-
Partnership ·
Samsung and Anthropic Partner to Bring Claude Enterprise AI to Galaxy Commercial Devices for APAC B2B
Samsung and Anthropic announce enterprise partnership integrating Claude AI capabilities into Samsung Galaxy commercial device programs — enabling APAC B2B customers in manufacturing, logistics, and financial services to deploy on-device and cloud-hybrid AI processing for Korean-language workflows, enterprise document analysis, and field operations AI on Samsung Galaxy commercial hardware.
-
Open source ·
ByteDance Open-Sources Doubao-1.5 Multilingual Model Family for APAC Enterprise Deployment
ByteDance releases Doubao-1.5 open-source model family under Apache 2.0 licence — 7B and 32B parameter variants trained with comprehensive Japanese, Korean, Mandarin Chinese, and Indonesian multilingual data, with APAC enterprise benchmark results showing superior performance versus Llama 3.1 on Asian-language reasoning, document understanding, and code generation tasks.
-
Regulation ·
Japan FSA Finalises AI Model Risk Management Framework for Financial Institutions
Japan's Financial Services Agency finalises AI model risk management framework requiring Japanese financial institutions to document model validation processes, report AI-related incidents within 48 hours, and conduct annual AI system audits — applying to AI-assisted credit scoring, algorithmic trading, fraud detection, and customer service AI deployed by Japanese banks, insurers, and securities firms.
-
Company ·
Kakao Corp Spins Out KakaoAI as Independent APAC Enterprise AI Subsidiary
Kakao Corp spins out KakaoAI as an independent APAC enterprise AI subsidiary — combining KakaoAI's Korean-English bilingual LLM with Kakao's 46 million South Korean users to offer enterprise AI services to Korean conglomerates expanding into Southeast Asian markets.
-
Security ·
CISA and APAC Agencies Publish Joint AI Security Guidance for Critical Infrastructure Operators
CISA and APAC cybersecurity agencies publish AI system security guidance for critical infrastructure — covering adversarial ML attack vectors, AI model supply chain risks, and incident reporting timelines for AI-enabled attacks on APAC energy, water, and transport systems.