GPT-5.5: Capabilities and Reactions

OpenAI's GPT-5.5, codenamed Spud, marks a significant leap in raw intelligence and agentic capabilities, positioning it as a direct competitor to Anthropic's Opus 4.7 for the first time in months. The model excels in well-specified coding tasks, computer use, and multi-step workflows, allowing users to delegate messy, multi-part tasks without micromanaging. OpenAI claims a 'much higher' intelligence level than GPT-5.4, with improved efficiency—using fewer tokens per Codex task while maintaining per-token latency. Pricing is set at $5 per million input tokens and $30 per million output tokens, with a 1 million token context window. However, real costs are lower due to token savings. Early reports from researchers highlight its ability to run experiment variations overnight from high-level algorithmic ideas, touching no code or terminal.

Despite the upgrade, GPT-5.5 isn't universally superior. For conversational tasks, exploratory work, or Claude Code-style projects, Opus 4.7 remains preferred. The model's strengths lie in structured, goal-oriented tasks like debugging, data analysis, and document creation. OpenAI hints at rapid iteration from here, suggesting future updates will focus on functionality rather than raw intelligence. The safety features have been strengthened, aiming to reduce misuse while preserving beneficial use. Overall, GPT-5.5 is a solid upgrade that splits effective usage with Anthropic's model, depending on the task nature.