GPT-5.5 and GPT-5.5 Pro land in Chat Completions and Responses API with 1M token context, built-in computer use, MCP, hosted shell, and tool search.
Key Takeaways
GPT-5.5 bundles computer use, hosted shell, apply patch, Skills, MCP, web search, tool search, and image input into a single API-accessible model.
GPT-5.5 Pro targets high-compute inference via the Responses API including Batch; pricing and rate limits not disclosed in this changelog entry.
GPT-5.5 defaults to medium reasoning effort, a regression from GPT-5.1’s none default – expect higher latency and cost at baseline.
Tool search, introduced in GPT-5.4, defers large tool surfaces to runtime to cut token usage, preserve cache hits, and reduce latency for tool-heavy agents.
GPT-5.5 Pro via Batch means async high-compute workflows are now possible without real-time API cost pressure.
Hacker News Comment Review
The release timeline raised a credibility flag: OpenAI publicly stated the day before that API deployment required additional safeguards, then shipped within roughly 24 hours – no explanation for the reversal was given.
Commenters surfaced genuine uncertainty about who the pro API tier is actually for; the use-case gap between pro API and pro ChatGPT is not obvious to practitioners.
Deepseek competitive pressure was offered as a casual explanation for accelerated release cadence, with no supporting evidence beyond timing intuition.
Notable Comments
@sigmoid10: Quoted yesterday’s safety statement verbatim and noted “one day counts as ‘very soon’” – direct contradiction between the two announcements with no changelog acknowledgment.
