OpenAI DevDay 2026: GPT-6.1 Sol, Agents API, and Codex Cloud

OpenAI used its DevDay 2026 showcase to roll out a broad set of developer-facing updates, and the through-line was unmistakable: the company is consolidating its stack around agents. Between a mid-tier model refresh, an Agents API that can now drive real software through graphical interfaces, and Codex environments untethered from the local machine, nearly every announcement served the same ambition of making autonomous, tool-using systems the default way developers build on top of OpenAI's platform 12.

GPT-6.1 Sol: the workhorse model gets an upgrade

The headline model release was GPT-6.1 Sol, an update to GPT-6 Sol that OpenAI positions squarely for coding, computer use, and professional tasks 2. The pricing story is the most notable part: OpenAI claims the model approaches GPT-6 Astra on several evaluations while charging one-fifth of Astra's standard input and output token prices, with cached input running at $0.10 per million tokens 2. That combination — near-flagship eval performance at a fraction of the cost — is aimed squarely at high-volume agent workloads, where token costs compound quickly across long, multi-step tasks.

The model is available through the API, ChatGPT Work, and Codex, meaning developers can put it to work in the same environments where the rest of the agent tooling lives 12.

The Agents API grows computer use

For developers building agents, the biggest functional change is computer use support in the Agents API, which lets applications operate software through graphical interfaces rather than just calling structured tools 2. The API also folds in multi-agent capabilities drawn from Codex, along with tool search, tool calling, and context compaction — with OpenAI managing the underlying execution infrastructure so teams don't have to stand up their own orchestration plumbing 2.

That last detail matters. By absorbing execution infrastructure into the platform, OpenAI is inviting developers to hand over a layer of the agent stack that many teams currently build or buy from third parties. The functionality is available through the API and inside Codex and ChatGPT Work for selected plans 2, though rollout comes with caveats worth watching — availability is staged, and architecture assumptions may differ across surfaces 1.

Codex moves to the cloud

Codex itself got a significant expansion: it can now run in cloud environments in addition to local machines, letting developers kick off remote tasks from other devices 2. That transforms Codex from a local developer companion into a distributed worker you can dispatch from anywhere.

The Codex CLI was updated in tandem, adding voice input and an /agents interface for delegating and monitoring multiple tasks at once 2. A new code-review workflow can analyze diffs and flag potential issues in GitHub pull requests and GitLab merge requests 2, extending Codex into the review stage of the development loop rather than just authorship.

On the security side, Codex Security Cloud can scan repositories and new commits, investigate findings, remove duplicates, and prepare fixes 2 — a full pipeline from detection to remediation, handled largely autonomously.

The quieter pieces: Decisions API, plugins, and Dots

Beyond the flagship items, the announcements included a Decisions API and new plugin capabilities for ChatGPT 12, along with a feature called Dots 1. These received less emphasis in coverage, but they point in the same direction: if agents are going to act on users' behalf, OpenAI needs standardized ways for them to make — and account for — decisions, and standardized ways for users to extend them with third-party functionality.

Why this matters

Taken together, DevDay 2026 reads less like a product demo and more like a repositioning of OpenAI's entire developer surface around agent infrastructure. The model tier is priced for agent economics; the Agents API handles execution so developers don't have to; Codex operates across cloud and local environments with review and security workflows attached; and plugin and decision frameworks supply the governance layer that agent deployments increasingly require 12.

The strategic implication for developers is that the moat question shifts. If OpenAI manages execution, review, and security scanning as platform services, much of the value in the middle of the agent stack — orchestration frameworks, glue code, bespoke infrastructure — gets absorbed. Teams building on this stack should weigh the convenience against the lock-in, and pay close attention to the staged rollout and architectural caveats that accompany these features in practice 1.

The bet OpenAI is making is that developers would rather rent the whole agent stack than assemble it. DevDay 2026 was the strongest signal yet that every future OpenAI release will be judged by whether it advances that bet.