Woofun AI reports that users detected a sharp reduction in the internal inference budget for GPT-5.6 Sol across all tiers, with the highest tier dropping from 960 to 128. This change sparked allegations that OpenAI had degraded the model to conserve computational resources.

OpenAI Codex lead Tibo Sottiaux clarified that the team adjusted inference intensity to investigate rapid subscription quota depletion. The experiment has been rolled back, confirming no permanent weakening. Officials noted that 372,000 contexts caused excessive quota deductions, prompting a temporary cap at 272,000.

Additionally, higher-than-expected Multi-Agent calls at high and xhigh tiers are being addressed. An inference optimization has been completed, granting all GPT-5.6 Sol subscribers an approximate 10% increase in usage.