OpenAI released GPT-5.5: stronger in programming, agents, and computer work

OpenAI presented GPT-5.5 — a new base model with explicit focus on programming, agent scenarios, and computer interaction. According to the company's description, this is not a cosmetic update, but a notable step forward in tasks where models need to not only respond, but also act more independently.

Where the model improved

The main focus of GPT-5.5 is on practical tasks where code accuracy, resilience in long chains of actions, and confident interaction with interfaces matter. OpenAI says the model writes code better, handles computer interaction more effectively, and is capable of performing more autonomous tasks without constant human oversight. For users, this means fewer manual fixes in typical work scenarios: from writing functions to completing multi-step processes.

OpenAI separately emphasizes that GPT-5.5 is the first fully retrained base model since GPT-4.5. This is an important signal: it's not just fine-tuned settings on top of an old foundation, but a deeper reworking of the model itself. This is indirectly confirmed by the growth in scores on agent programming benchmarks. It's in such benchmarks that one can see how well a model maintains context, plans steps, and completes a task to the end, rather than simply generating plausible text.

Where it's available

The release has already been rolled out to ChatGPT and Codex users, but access depends on the plan. The base GPT-5.5 has appeared in Plus, Pro, Business, and Enterprise plans. For those who need maximum performance, OpenAI is separately preparing GPT-5.5 Pro — this version is intended for Pro, Business, and Enterprise. That is, the company immediately differentiates the mass-market and more expensive professional use scenario.

In practice, this looks like:

• GPT-5.5 is already available in ChatGPT
• The model is also in Codex
• Plus, Pro, Business, and Enterprise plans are supported
• GPT-5.5 Pro version is separate
• API access for developers is currently marked as coming soon

This rollout order shows OpenAI's familiar strategy: first give the model to end users within its own products, then open it via API for external teams and services. For developers, this means a pause: testing the new model in live integrations right now is not possible, even though the release already generates interest precisely in code and agent scenarios. Until then, comparing the model is only possible within OpenAI's interfaces.

Price and expectations

The most controversial aspect is cost. The price per token for GPT-5.5 is twice that of GPT-5.4. For companies that actively process large volumes of requests, this is not a cosmetic difference, but a direct recalculation of the product's economics. If the new model is indeed significantly better at programming and autonomous work, part of the market will accept the price increase. But if the improvement turns out to be not as large in real-world tasks, the price discussion will quickly move to the forefront.

This is the first fully retrained base model since GPT-4.5.

This phrasing sets the main tone of the release. GPT-5.5 is presented as an update that should be felt in work, not just in name. Therefore, attention to it will be particularly high among developers, teams building agent products, and companies using models in partially autonomous processes. They will be the first to verify whether the promises of improved code, computer interaction, and autonomy hold up outside of demos and tests.

For now, the picture is mixed: on one hand, the release looks stronger than a typical minor update and could become an important milestone after GPT-5.4. On the other — full assessment will only begin after the API appears and the first independent comparisons by price, quality, and stability in real pipelines. It's this stage that will show whether we're talking about a real shift or just an expensive improvement on paper for the market.

What this means

GPT-5.5 shows that OpenAI is again shifting focus from chat for chat's sake to models that should deliver measurable value in code, agents, and workflows. If the quality improvement is confirmed in practice, the market will get not just a new version, but a more suitable tool for autonomous AI scenarios — albeit at a notably higher price. For product teams, this is a reason to recalculate where a more expensive model pays for itself.