Introducing GPT-5.5

Source: OpenAI Date: 2026-04-23 URL: https://openai.com/index/introducing-gpt-5-5

Summary

OpenAI’s launch announcement for GPT-5.5, internally codenamed “Spud” through development. The model ships positioned as the new frontier-tier offering, with the major framing being benchmark surface replacement rather than incremental improvement on the existing ladder — OpenAI is signaling that the prior benchmark suite has saturated, and 5.5 is the model that exposes the next set of evaluations. Pricing, availability, and exact capability claims are gated behind the launch page (the source page returns 403 to direct fetches; details were carried by secondary coverage and the daily reports cohort).

(Source page is paywalled to direct read; this signal stub captures the event itself. Detailed analysis lives in the corresponding daily report and the GPT-5.5 thread on /landscape/threads.)

Implications

Three threads this signal feeds:

The benchmark-saturation thread. “Spud” being framed as the model that replaces a benchmark ladder rather than topping one is itself a category move. OpenAI is signaling that the existing eval surface has run out, which puts pressure on every vendor that’s been competing on those scores. Anthropic’s Opus 4.7 (“less broadly capable” positioning vs. Mythos), DeepSeek V4 (largest open-weight, MIT), and Kimi K2.6 (agent swarm framing) all have to recalibrate against a new yardstick.

Cadence pressure. GPT-5.5 lands the same week as Opus 4.7, Kimi K2.6, DeepSeek V4. Four major model releases within ~10 days means the integrators (Cursor, Aider, OpenCode, Codex CLI itself) are integrating against a moving target. Watch whether agent products freeze on a single model for stability or scramble per-release.

OpenAI-as-platform. Three companion launches — System Card, Automations, Top 10 uses for Codex at work — show OpenAI is shipping product context alongside the model, not just weights. That’s the Anthropic playbook (Managed Agents, Skills, Channels) being mirrored from the other side.

Watch:

• Pricing tier, particularly relative to Opus 4.7’s $5/$25 and Mythos
• Whether GPT-5.5 becomes the default in Codex CLI immediately or stays opt-in
• The specific benchmarks the new surface introduces — these define what 5.5’s competitors will need to publish
• Adoption signals from secondary CLIs (Aider, OpenCode, Vibe) within ~2 weeks