OpenAI DevDay 2026: GPT-6.1 Sol, Computer-Use Agents and Cloud Codex Headline Developer Push

OpenAI packed its DevDay 2026 keynote with developer-facing releases, led by a new flagship model it calls GPT-6.1 Sol. The company also added computer use to its Agents API, introduced cloud-hosted Codex environments, rolled out a Decisions API, and expanded plugin support inside ChatGPT. The batch, reported by InfoQ's Daniel Dominguez, points to a platform strategy that now spans models, agent tooling, and an app layer rather than a single API.

GPT-6.1 Sol is the headline item and the one most teams will evaluate first. OpenAI has not yet published full benchmark tables in the material summarized here, so the practical question for engineering leads is how Sol performs against the prior generation on their own evals and whether pricing and latency shift with it. Until those numbers are public, adoption decisions will rest on internal testing rather than vendor charts.

Computer use for the Agents API is the second structural change. Letting agents drive a graphical interface widens what automation can touch, from legacy internal tools to third-party sites that never shipped an API. That reach also raises the stakes on sandboxing, permission scoping, and audit logging, since an agent that can click can also click the wrong thing at scale.

Cloud-based Codex environments move the coding assistant off the local machine and into infrastructure OpenAI manages. For teams, that means shared, reproducible setups instead of per-developer configuration, which matters for onboarding and for keeping agent runs consistent across a team. It also concentrates more of the development loop inside OpenAI's platform, a trade-off security and platform teams will want to weigh.

The Decisions API and the new ChatGPT plugin capabilities round out the release. A Decisions API suggests OpenAI wants a first-class primitive for routing or policy logic inside applications, while plugins give third parties a distribution channel into ChatGPT's user base. Together they push OpenAI further into territory occupied by orchestration frameworks and app marketplaces alike.

What to watch next is the fine print: published Sol benchmarks, computer-use pricing and rate limits, Codex cloud's data-retention terms, and which partners ship plugins on day one. OpenAI has set the agenda for its developer base with this DevDay slate. The follow-through on documentation, quotas, and security controls will decide how quickly teams move from demos to production.