Users report that GPT-5.5 Codex displays degraded performance on complex tasks, potentially linked to reasoning-token clustering. According to GitHub issue #30364, this clustering appears to occur at specific intervals of 516, 1034, and 1552 tokens. The issue is officially classified within the OpenAI Codex repository under three specific labels: bug, model-behavior, and rate-limits. These labels indicate that the observed problems involve unexpected model behaviors, potential token usage reporting errors, and concerns regarding system rate limits. As of July 4, 2026, the discussion regarding these performance issues has garnered significant attention from the community, including an active thread on Hacker News with 65 points and 8 comments.

Users report that GPT-5.5 Codex displays degraded performance on complex tasks, potentially linked to reasoning-token clustering. According to GitHub issue #30364, this clustering appears to occur at specific intervals of 516, 1034, and 1552 tokens. The issue is officially classified within the OpenAI Codex repository under three specific labels: bug, model-behavior, and rate-limits. These labels indicate that the observed problems involve unexpected model behaviors, potential token usage reporting errors, and concerns regarding system rate limits. As of July 4, 2026, the discussion regarding these performance issues has garnered significant attention from the community, including an active thread on Hacker News with 65 points and 8 comments.