Microsoft is readying its servers for GPT-4.5 and GPT-5

Microsoft is clearing the runway for OpenAI’s next frontier.

According to The Verge, Microsoft engineers are currently preparing server capacity to host both GPT-4.5 and the highly anticipated GPT-5. People familiar with the plans say GPT-4.5 could land on Microsoft servers as early as next week, while GPT-5 is currently slated for late May.

Killing the model picker

GPT-4.5, codenamed Orion, represents OpenAI’s last traditional, non-chain-of-thought frontier model before the architecture shifts entirely. It was originally expected by late 2024 but slipped into early this year.

The much bigger shakeup is GPT-5. Sam Altman has described it not just as a single raw checkpoint, but as a unified system integrating multiple breakthroughs. Most notably, OpenAI is no longer planning to ship its flagship o3 reasoning model as a standalone release. Instead, o3 will be baked directly into the GPT-5 system.

The end goal is eliminating ChatGPT’s cluttered dropdown menu. Altman recently posted on X that the team hates the model picker just as much as users do, aiming to return to a seamless unified intelligence that automatically knows when to route queries to deep reasoning.

Headed for a showdown at Build

Timing GPT-5 for late May places it squarely against Microsoft’s Build developer conference, kicking off May 19th. That event directly overlaps with Google I/O that same week, setting the stage for massive competitive fireworks.

Unlike last year—when OpenAI’s surprise launch of GPT-4o caught Microsoft off guard and undercut some paid Azure features—Redmond is heavily prepared this time around. Microsoft is revamping Copilot to match the unified approach and is actively building its own web-navigating agent to rival OpenAI’s Operator, with a major internal push to slash enterprise costs.

Why it matters

OpenAI and Microsoft are finally fixing AI’s biggest usability headache: forcing users to guess which engine to run. Folding o3 directly into GPT-5 rather than shipping it as another niche tool is the right move for mainstream adoption. Nobody outside of AI Twitter wants to deliberate whether a quick spreadsheet edit needs an analytical reasoning engine or a standard chat model.

However, routing automation introduces real trade-offs. If the unified system misjudges user intent, developers and power users could face unpredictable token pricing and wild swings in latency. For Microsoft, locking down server capacity months ahead is defensive necessity. With Google I/O running side by side with Build, the pressure to deliver a unified, cost-effective enterprise agent ecosystem has never been higher.