OpenAI and Meta have entered the competitive business AI market by unveiling next-generation artificial intelligence (AI) models and automation tools. The focus has shifted to securing AI agents capable of performing practical tasks such as scheduling, document and spreadsheet work, and coding.
OpenAI Unveils GPT-5.6 and Business Agent 'ChatGPT Work'
OpenAI has publicly released its latest AI model, 'GPT-5.6', after undergoing a pre-review process by the U.S. government. Following a limited release to select institutions at the end of last month, the model is now available to general users. OpenAI conducted a restricted pre-release and additional review at the government's request before fully deploying the model.
The GPT-5.6 product line includes three models: the top-tier 'Sol', the balanced 'Terra', and the cost-effective 'Luna'. According to evaluation results released by OpenAI, GPT-5.6 outperformed some competitors, including Anthropic's Claude 5, in several areas. The top model, Sol, demonstrated performance comparable to Anthropic's advanced model, Claude Mythos 5.
Additionally, OpenAI has launched 'ChatGPT Work', a business AI agent that combines ChatGPT with the coding agent Codex. ChatGPT Work is designed to create documents, presentations, spreadsheets, and web applications. It can gather information from specified applications and files, breaking down complex tasks into manageable steps. This positions it in competition with other business AI agents, such as Anthropic's 'Claude Co-Work' and Microsoft's 'Copilot Co-Work'.
Meta Releases Coding AI 'Muse Spark 1.1' for Enterprises
On July 9, Meta announced the release of its multimodal, agent-based coding model, 'Muse Spark 1.1'. Previously, Muse Spark was accessible only to select partners through a private API preview. However, Muse Spark 1.1 is now available in a public API preview format, allowing new users to access it. Businesses will pay based on token usage, marking the first time Meta has offered an AI model for enterprise use with a fee structure. According to Meta, internal developers and researchers are already utilizing Muse Spark 1.1 for their work.
Muse Spark 1.1 supports a context window of up to 1 million tokens, enabling it to reuse previous actions and information throughout lengthy tasks while retaining only essential context. It also features multimodal capabilities, allowing it to understand images, videos, and audio. This enables the generation of code based on visual information and detailed descriptions of images and videos, linking recognition results to actual tasks.
Meta Invests CAD 13 Billion in First AI Data Center in Canada
Meta is set to build its first data center in Canada, located in Sturgeon County, Alberta. The company announced via its blog that it will construct a 1 gigawatt (GW) data center in Alberta, with a total investment of CAD 13 billion (approximately KRW 13.7 trillion). This facility will be developed as a 1 GW infrastructure for AI computations. The new data center will be the 33rd facility built by Meta, joining existing centers in Ireland, Sweden, and Singapore. Earlier, Meta CEO Mark Zuckerberg indicated that the company is considering entering the cloud business by selling excess computing capacity generated at the data center or providing access to its AI models.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.