OpenAI is sharpening its enterprise AI strategy with the launch of ChatGPT Work, a new agentic platform designed to automate workplace tasks, alongside the broader rollout of its GPT-5.6 models, which the company says deliver stronger performance at lower operating costs.
According to the company, ChatGPT Work can operate across applications and files, execute long-running tasks, coordinate multiple tools, and produce business documents, presentations, spreadsheets, and websites, allowing employees to delegate more complex workflows rather than interact through individual prompts.
GPT- 5.6 models, generally available weeks after a . “The result is stronger performance per dollar: more successful work for the same spend, or comparable results at a lower total cost.”
The models are now generally available through ChatGPT, Codex, and the OpenAI API. OpenAI has priced Sol at $5 per million input tokens and $30 per million output tokens, while Terra and Luna provide progressively lower-cost options for organizations scaling AI deployments, the statement added.
Enterprise AI shifts from experimentation to economics
ChatGPT Work combines GPT-5.6 with enterprise integrations and agentic capabilities that allow users to perform multi-step tasks across connected business applications instead of interacting with AI through isolated prompts. OpenAI said the platform is designed to help organizations automate knowledge work while maintaining enterprise-grade governance and security.
The launch comes as enterprises move beyond AI experimentation and begin deploying models across production workloads, making inference costs a growing concern for CIOs.
“The AI wave has brought productivity gains, but rising token consumption has also created bill shocks for enterprises,” said Neil Shah, vice president for research and partner at Counterpoint Research. “This is forcing organizations to adopt different models for different workloads, making performance per dollar the key metric.”
Faisal Kawoosa, co-founder and chief analyst at Techarc, said enterprises are now evaluating AI investments more pragmatically.
“The exploratory stage of AI is over,” he said. “Organizations can derive value from AI today, but performance per dollar will determine whether it becomes part of everyday business operations or remains an ad hoc tool.”
Tiered models for different workloads
GPT-5.6 Sol is OpenAI’s flagship model for complex reasoning, Terra targets mainstream enterprise applications, and Luna is designed for lower-cost, high-volume deployments.
According to OpenAI, GPT-5.6 Sol scored 53.6 on Agents’ Last Exam, a benchmark for long-running professional workflows, outperforming competing frontier models while requiring significantly lower compute costs.
The company also introduced two new reasoning modes. The max mode allocates additional compute for complex problems, while ultra coordinates four AI agents in parallel to accelerate demanding workflows.
“Ultra goes further by coordinating four agents in parallel by default, trading higher token use for stronger results and faster time-to-result on demanding tasks,” the statement added.
Shah said the architecture reflects how enterprises are increasingly orchestrating multiple AI models.
“GPT-5.6 gives enterprise architects flexibility to route workloads from Luna to Sol depending on whether they require automation, logic, or complex reasoning,” he said.
Kawoosa added that the tiered approach aligns with how enterprise software has traditionally been consumed.
“It gives enterprises of different sizes the flexibility to optimize technology consumption according to their requirements,” he said.
Coding, productivity, and security gains
OpenAI said GPT-5.6 Sol achieved a score of 80 on the Artificial Analysis Coding Agent Index while consuming fewer than half the output tokens of competing models. It also reported state-of-the-art results on Terminal-Bench 2.1 and DeepSWE, benchmarks that measure real-world software engineering tasks.
The company said the models also improve enterprise productivity through stronger document generation capabilities and integrations with Microsoft 365, Google Drive, Slack, and Notion.
On cybersecurity, GPT-5.6 Sol scored 73.5% on ExploitBench, up from 47.9% for GPT-5.5, and nearly doubled its predecessor’s performance on ExploitGym.
“GPT-5.6 supports important defensive tasks such as secure code review, patching, threat modeling, and blue teaming,” OpenAI said.
Security remains an enterprise focus
OpenAI said GPT-5.6 incorporates its “most robust safeguards to date,” combining model-level protections with real-time monitoring and extensive safety testing, including approximately 700,000 GPU hours of automated red-team evaluations.
Shah said layered guardrails and monitoring could become an important differentiator for enterprise deployments.
Kawoosa, however, said CIOs will continue demanding greater transparency before fully trusting frontier AI systems.
“Competition among LLM providers will continue, with vendors constantly testing and challenging each other’s guardrails,” he said.
The article originally appeared on InfoWorld.
SOCIAL SHARE CARD GENERATOR