OpenAI has announced the launch of its latest AI model called GPT-5.6. The new family of models includes the new flagship model Sol, alongside Terra, a balanced model for everyday work, and Luna, OpenAI’s most cost-efficient model. It also launched ChatGPT Work, “an agent that can take action across your apps and files, stay with a project for hours if needed, and turn a goal into finished work,” per OpenAI.
GPT-5.6: Everything to Know
OpenAI has unveiled GPT-5.6, its latest family of AI models, claiming major improvements in coding, reasoning, knowledge work, cybersecurity, and scientific research while delivering better performance at lower cost than previous generations. The new lineup includes three models: GPT-5.6 Sol, Terra, and Luna, each targeting different performance and pricing tiers.
GPT-5.6 Sol serves as the flagship model in the lineup. According to OpenAI, it can complete more tasks using fewer tokens, which reduces overall inference costs alongside maintaining higher accuracy. The company says the model outperforms competing frontier models across several benchmarks, including long-running professional workflows, coding, scientific reasoning, and agentic tasks.
For developers, GPT-5.6 brings improvements to coding performance. OpenAI says Sol sets a new high score on the Artificial Analysis Coding Agent Index while using fewer output tokens, completing tasks faster, and costing less than competing models. The model also posts state-of-the-art results on benchmarks such as Terminal-Bench 2.1 and DeepSWE, which evaluate command-line workflows and software engineering tasks. OpenAI has also expanded Programmatic Tool Calling in the Responses API, allowing GPT-5.6 to write and execute lightweight programs that coordinate tools, process intermediate results, and automate multi-step workflows more efficiently.
The company is also introducing two new reasoning modes for demanding workloads. Max gives GPT-5.6 additional time to analyse problems, explore alternatives, and verify its work before responding. Ultra goes a step further by coordinating four AI agents in parallel by default, enabling faster completion of complex tasks while improving overall accuracy. Developers can also build similar multi-agent workflows through the beta version of the Responses API.
OpenAI says GPT-5.6 also represents a significant upgrade for design and creative work. The model is better at generating polished interfaces from natural language prompts and can inspect rendered outputs to identify visual or functional issues before refining them. It can also create interactive visualisations, presentations, and editable documents with improved layouts, typography, and formatting.
Knowledge work is another major focus for GPT-5.6. The model is designed to work with documents and data from platforms such as Slack, Notion, Microsoft 365, and Google Drive, transforming raw information into polished reports, presentations, spreadsheets, and other professional outputs. OpenAI claims the model performs better when following existing templates, maintaining consistent formatting, and generating more accurate financial models and technical documents.
In cybersecurity, GPT-5.6 introduces improvements in secure code review, vulnerability analysis, patch validation, and threat modelling. OpenAI says the model achieves significantly higher scores than GPT-5.5 across several security benchmarks while remaining below its internal “Critical” capability threshold. The company is making advanced cybersecurity capabilities available through its Trusted Access programme for verified researchers and organisations working on defensive security.
The new model family also delivers gains in scientific research. OpenAI says GPT-5.6 performs better across biology, chemistry, and life sciences benchmarks, while internally the company is already using it to accelerate AI research, debug training systems, optimise experiments, and improve machine learning workflows.
Alongside the performance upgrades, OpenAI says GPT-5.6 introduces its most comprehensive safety system to date. The company combined large-scale automated testing with extensive human red teaming before launch and added layered safeguards that include model-level protections, real-time monitoring, reasoning-based safety checks, and account-level enforcement. OpenAI says the new safeguards block around ten times more potentially harmful cybersecurity-related activity compared to previous models while continuing to support legitimate defensive work.
GPT-5.6 is available starting today across ChatGPT, Codex, and the OpenAI API. The rollout is taking place globally and is expected to reach all users over the next 24 hours. Plus, Pro, Business, and Enterprise subscribers can access GPT-5.6 Sol through higher effort settings, while Free and Go users get access to GPT-5.6 Terra. Developers can access all three models through the OpenAI API.
OpenAI has also announced pricing for the new models. GPT-5.6 Sol costs $5 per million input tokens and $30 per million output tokens. Terra is priced at $2.50 for input and $15 for output per million tokens, and then Luna costs $1 for input and $6 for output per million tokens. The company has also introduced more predictable prompt caching, including support for explicit cache breakpoints and a minimum cache lifetime of 30 minutes.
Also Read: OpenAI Launches GPT-Live Voice Model in ChatGPT
ChatGPT Work: Everything to Know
OpenAI has introduced ChatGPT Work, a new agentic workspace within ChatGPT that is designed to handle complex, long-running tasks across connected apps and workflows. Powered by the newly announced GPT-5.6 model and built on Codex technology, ChatGPT Work can gather information from multiple sources, create finished deliverables such as documents, spreadsheets, presentations, websites, and web apps, and independently complete multi-step workflows over several hours.
Unlike traditional chat interactions, ChatGPT Work can break larger projects into smaller tasks, execute them independently, and keep working even when users are away. It supports tasks such as analysing financial reports, preparing sales presentations, creating marketing campaign briefs, and generating business documents from connected enterprise data. Users can monitor progress, answer follow-up questions, redirect workflows, and approve important actions whenever required.
OpenAI says ChatGPT Work can also automate entire workflows. For example, it can transform customer research into a campaign brief, generate marketing assets based on that brief, and localise the content for different regions while maintaining context throughout the process. Scheduled Tasks further extend these capabilities by allowing ChatGPT Work to automatically perform recurring actions, such as updating documents from Slack or Microsoft Teams conversations, monitoring websites for changes, summarising new information, and more.
ChatGPT Work is tightly integrated with plugins that connect it to workplace tools such as Slack, Microsoft Teams, Google Drive, SharePoint, Gmail, Outlook, calendars, CRM platforms, project management software, and other enterprise applications. Users can reference connected services directly in prompts using the “@” symbol, while ChatGPT can also automatically determine when to pull relevant information from connected apps.
Alongside ChatGPT Work, OpenAI has introduced Sites in public beta. Sites allow users to generate interactive websites and web applications directly within ChatGPT. These can be shared through a URL and are intended for use cases such as live dashboards, project trackers, launch calendars, internal portals, prototypes, and interactive reports. The generated sites can also update automatically as underlying data changes.
The ChatGPT desktop application has also received significant upgrades. It now includes a built-in Browser that enables ChatGPT to gather information from websites, interact with web-based tools, and work directly with files stored in Google Workspace and Microsoft 365. On desktop, Computer Use allows ChatGPT to interact with local applications by clicking, typing, opening files, and completing workflows across software installed on the user’s computer. OpenAI says these actions can be performed either on demand or as part of Scheduled Tasks.
The existing Codex desktop application is being merged into the new ChatGPT desktop app. Codex retains its developer-focused capabilities while gaining new features such as inline code editing, pull request reviews, improved computer use powered by GPT-5.6, and support for multiple repositories within a single project. OpenAI also announced that its Atlas browser will be phased out as similar browser capabilities are now integrated directly into ChatGPT.
For enterprise customers, ChatGPT Work builds on the security, privacy, compliance, and administrative controls already available in ChatGPT Enterprise. Administrators can manage plugin access, browser permissions, connected applications, local computer actions, and also workspace-level policies. Then there’s Auto-review which further adds another layer of protection by reviewing important actions involving connected systems before they are executed, which helps reduce the risk of accidental exposure of sensitive information.
ChatGPT Work begins rolling out today on the web and mobile apps for Pro, Enterprise, and Edu subscribers, with Plus and Business users receiving access over the next few days. The updated ChatGPT desktop application is available globally for both Windows and macOS and includes Chat, Work, and Codex modes across all plans, including the Free tier. Existing Codex users can simply update the application, while the previous ChatGPT desktop App will continue to be available under the name ChatGPT Classic.
OpenAI notes that ChatGPT Work follows a different usage model than standard chat conversations. Since tasks can run for extended periods and consume more computing resources, usage depends on the complexity and duration of the work performed. Enterprise and Edu administrators can also configure spending controls and workspace-level usage limits through the Admin Console as adoption increases.
Author: The Mobile Indian Network
Source: The Mobile Indian
Reviewed By: Editorial Team
