OpenAI's GPT-5.5 lands — same speed as 5.4, smarter on everything

🚨 The Big News

OpenAI's GPT-5.5 lands — same speed as 5.4, smarter on everything

OpenAI shipped GPT-5.5 today, codenamed "Spud" internally. It's a step up from GPT-5.4 across coding, computer use, and deep research — and it's matching 5.4's per-token latency while doing it. That's the part worth pausing on. Bigger models usually mean slower serving, but OpenAI claims they kept the same speed while bumping the intelligence.

What's new:

• Better at writing and debugging code, online research, data analysis, and document/spreadsheet creation
• Improved at operating software and chaining tools together until a task is done
• Same per-token latency as GPT-5.4 in real-world serving — you don't pay a speed tax for the upgrade
• Rolling out to Plus, Pro, Business, and Enterprise users in ChatGPT and Codex starting today; API access via gpt-5.5 and gpt-5.5-pro follows on April 24
• Strongest safety guardrails OpenAI has shipped, with new Bio Bug Bounty program for security testing

Why this matters: This is OpenAI's response to Claude Opus 4.7 (which dropped a week ago) and the broader push toward AI that finishes whole tasks autonomously rather than answering one prompt at a time. If you're building agent workflows, GPT-5.5 narrows the gap with Claude on coding while keeping OpenAI's lead in computer use. For most ChatGPT users, you'll notice it as "Spud just feels like it gets it" without thinking about why.

Source →

Related drops from OpenAI today

🔑 New · Safety GPT-5.5 System Card published The full safety evaluation document went up alongside the model launch. Covers preparedness framework results, internal and external red-teaming, and targeted testing for advanced cybersecurity and biology capabilities. Worth a skim if you care about what guardrails actually got tested.

Source →

🛠️ New · Security program GPT-5.5 Bio Bug Bounty launches OpenAI opened a dedicated bounty program for finding bio-related safety issues in GPT-5.5. If you're a security researcher with bio domain expertise, this is the formal path to get paid for responsible disclosure.

Source →

The AI ChangeLog is generated daily by Demand Signals. We scrape official news + changelogs, run them through Claude, and publish a plain-English summary so you don't have to read the docs. Subscribe to our blog for daily updates.