OpenAI readies GPT-5.1 Thinking model ahead of Gemini 3 Pro (www.testingcatalog.com)

OpenAI appears to be preparing a GPT-5.1 family, with a variant labeled "GPT-5.1 Thinking" showing up in ChatGPT’s web backend — a strong signal an update is imminent. The Thinking model is described as handling advanced reasoning, suggesting optimizations for multi-step reasoning and agent-like workflows. Reports also point to a broader lineup (Mini, Thinking, and a possible Codex-focused upgrade) seen in internal test environments under obfuscated names and A/B test setups, implying OpenAI is tuning multiple configurations for different latency, cost, and capability trade-offs. Technically, leaks reference multiple models with differing "thinking budgets," likely meaning variations in context window size, compute allocation per request, or depth of reasoning, which would shape consumer and enterprise offerings. The timing likely reflects competitive pressure from Google’s upcoming Gemini 3 Pro; both vendors are racing to deliver larger context windows and more reliable, nuanced problem solving. If realized, these changes could materially improve developer tooling, agent behaviors, and high-context enterprise workflows — though availability, tiering, API access, and benchmark performance remain unconfirmed. The discovery came from inspecting web app components; TestingCatalog and others are watching for official specs and performance data as the launch window nears.