OpenAI launches GPT-5.5 with agentic coding and research upgrades

At a glance:

• OpenAI has released GPT-5.5, an upgrade to the models powering ChatGPT and Codex, with claimed gains in multi-step work, tool use, and self-verification.
• Two variants are available at launch: GPT-5.5 Thinking for faster help on harder problems, and GPT-5.5 Pro positioned as a research partner where accuracy matters more than speed.
• Access is tiered by plan: ChatGPT Plus, Pro, Business, and Enterprise get GPT-5.5 Thinking; GPT-5.5 Pro is limited to ChatGPT Pro, Business, and Enterprise; Codex spans Plus, Pro, Business, Enterprise, Edu, and Go, with API access said to be coming very soon.

Model strategy and capabilities

OpenAI describes GPT-5.5 as better at multi-step work, claiming it can plan, use tools, and verify its own output with less hand-holding. The model is said to offer gains in agentic coding, computer use, and early-stage scientific research, positioning it as a step toward more autonomous workflows inside ChatGPT and Codex. OpenAI argues that its latest model is more token-efficient, so Codex tasks should — in theory — finish with less overhead despite the capability bump, which could change cost and latency profiles for heavy users over time.

The split into GPT-5.5 Thinking and GPT-5.5 Pro reflects a deliberate product segmentation. GPT-5.5 Thinking offers faster help for harder problems, according to OpenAI, while GPT-5.5 Pro is being pitched as a research partner for tougher questions where accuracy matters more than speed. This mirrors broader industry moves to specialize models for latency-sensitive interactions versus deep reasoning, and it gives teams a clearer choice between throughput and precision without swapping apps or re-architecting prompts.

Access, pricing, and rollout

Access to GPT-5.5 is gated by subscription tier and use case. ChatGPT Plus, Pro, Business, and Enterprise subscribers get GPT-5.5 Thinking, while the more powerful GPT-5.5 Pro model is limited to ChatGPT Pro, Business, and Enterprise. In Codex, GPT-5.5 spans Plus, Pro, Business, Enterprise, Edu, and Go plans, widening availability for developers and students relative to prior flagship releases. API access is said to be coming very soon, though OpenAI has not provided dates, regions, or pricing details for programmatic use.

The tiering intersects with OpenAI's recent pricing moves. Earlier in April 2026, OpenAI added a new $100/month ChatGPT subscription tier meant to support increasing Codex use, providing 5x more Codex usage than the $20/month ChatGPT Plus plan and targeting longer, high-effort Codex sessions. That plan competes directly with Anthropic's Claude Code, and the arrival of GPT-5.5 could shift the value equation for teams weighing speed, token efficiency, and accuracy across long coding or research tasks.

Ecosystem and interface expansions

OpenAI has been widening ChatGPT's footprint beyond the browser and mobile apps. In late March 2026, OpenAI updated ChatGPT with support for CarPlay, allowing CarPlay users to ask ChatGPT questions and make requests directly from their vehicle dashboard after Apple began permitting third-party voice-based conversational apps to interface with CarPlay in iOS 26.4, provided apps implement the feature and obtain a special entitlement. This expands the contexts in which GPT-5.5 could be invoked hands-free, from coding assistants to in-car research and task management.

At the same time, competitive dynamics are sharpening. Google recently launched a Gemini Import Tool to make it easier for customers to switch from ChatGPT, Claude, and other AI apps, promising to import memories, context, and chat history so Gemini can understand a user's preferences, relationships, and personal context. As OpenAI pushes deeper into agentic workflows with GPT-5.5, retention and switching costs become more salient, and the ability to move context across platforms could influence how teams adopt or abandon agentic stacks.

Implications for coding and research workflows

GPT-5.5's focus on agentic coding and computer use suggests OpenAI is betting on models that can orchestrate multi-step tasks with less human steering. If token-efficiency claims hold, Codex sessions could become cheaper and faster even as capabilities rise, potentially altering how businesses budget for AI-assisted development and research. Early-stage scientific research gains are harder to quantify without benchmarks or peer review, but the stated direction — planning, tool use, and self-verification — points to models that can draft, test, and refine hypotheses with lighter supervision.

The rollout also raises questions about access and equity. By restricting GPT-5.5 Pro to higher-tier plans and limiting Codex availability on lower tiers, OpenAI is effectively segmenting the market between casual users and power developers or research teams. API access will be the next inflection point: once endpoints open, third-party tools and enterprises can integrate GPT-5.5 into custom pipelines, and pricing and rate limits will determine whether token-efficiency gains translate into real cost savings at scale.

What to watch next

Key unknowns include API release timing, regional availability, and pricing, as well as independent evaluations of GPT-5.5's reasoning and tool-use claims. The interplay between GPT-5.5 and OpenAI's newer $100/month tier will clarify whether token efficiency and capability gains justify higher spend for Codex-heavy workflows, and how that stacks up against Claude Code and Gemini in practice. As agentic coding matures, expect metrics beyond raw accuracy — such as plan success rates, tool-use reliability, and human-in-the-loop overhead — to become central to vendor comparisons and procurement decisions.