SambaNova Systems, the generative AI solutions company with the fastest models and most advanced chips, announces that “Typhoon”, a Thai Large Language Model (LLM) powered by SCBX and SCB 10X, has been introduced into the Samba-1 Composition of Experts (CoE) LLM.
Mr. Kasima Tharnpipitchai, Head of AI Strategy at SCB 10X, a subsidiary of SCBX Group, stated: “We are pleased that Typhoon is now a part of Samba-1, making it available to SambaNova’s customers around the world. The power and scalability of Samba-1 is now enhanced with Typhoon, a series of Thai Large Language Models (LLMs).”
Typhoon-1.5X has been recently launched, offering significant advancements in Thai NLP and AI. Available in 8B and 70B sizes, it is optimized for the Thai language and performs comparably to top models like ChatGPT and GPT-4. Built on Llama3, it features enhanced training and alignment techniques, including improved human alignment in instruct models.
Typhoon is now available in open beta, offering two models. The Pre-trained Model focuses on Thai language proficiency, covering vocabulary, context, cultural nuances, and global knowledge. This model is open source and freely available for download, promoting competitiveness in Thailand’s AI industry. The Instruction-tuned Model builds on this foundation, enabling tasks such as translations, summaries, and question answering, and is accessible through an API service.
Mr. Kaweewut Temphuwapat, Head of R&D and Innovation Lab, SCBX noted that “large language models have emerged as powerful tools, and Typhoon has the potential to truly lead the Thai AI landscape. It addresses the language gap in models primarily trained in English by overcoming resource limitations associated with the Thai language. By integrating Typhoon into Samba-1, we believe AI developers will have an enhanced experience, advancing AI innovations and strengthening Thailand’s AI industry.”
“Samba-1 offers the most advanced models from open source, which ensures that we always have access to the world’s best and fastest AI models,” said Rodrigo Liang, Co-Founder and CEO of SambaNova Systems. “The addition of Typhoon fits into this strategy; we’re pleased to incorporate its capabilities into this world-leading model.”
“AI developers across industries including banking and financial services, now have a new platform to accelerate their AI journey,” said Marshall Choy, SVP Product, SambaNova Systems. “The integration of Typhoon LLMs into the Samba-1 Composition of Experts delivers highest inference performance with less infrastructure and cost than ever before.”
SambaNova’s unique Composition of Experts architecture aggregates multiple expert models and improves performance and accuracy by selecting the best expert for each application. Typhoon runs optimally on SambaNova’s SN40L chip with its 3-tier Dataflow memory architecture.