What's Happening?
Users of ChatGPT Plus are reporting significant performance issues with Codex, specifically experiencing severe latency and what they describe as an extremely limited usage quota. According to a user in the OpenAI Developer Community, the output generation
speed for Codex averages only about 37 tokens per second, a substantial drop from the approximately 250 tokens per second observed with competing models like Gemini and DeepSeek on the same network. This performance degradation, which began around November 16, 2026, severely impacts daily development and coding efficiency. Additionally, the usage quota for Codex under ChatGPT Plus is perceived as overly restrictive, estimated to be roughly one-tenth of Gemini's $10 tier and one-fifth of Claude's $20 tier, further hindering the coding workflow.
Why It's Important?
These issues are critical for U.S. developers, businesses, and individuals who rely on ChatGPT Plus and Codex for their coding and development tasks. The severe drop in generation speed directly translates to decreased productivity and increased time spent on projects, potentially leading to higher operational costs for businesses. The restrictive usage quotas further exacerbate this problem, limiting the utility of a paid subscription service. This situation could force developers and companies to seek alternative AI coding assistants, impacting OpenAI's market position and revenue from its premium services. It also highlights the challenges in maintaining consistent performance and value proposition in the rapidly evolving AI tools market.
What's Next?
Users are urging the OpenAI Team to investigate the throughput issues affecting Codex models and to consider adjusting the Plus tier limits. The user's troubleshooting steps, including starting fresh sessions and verifying network connections, suggest the problem is likely server-side, possibly due to throttling, rate limiting, or routing bottlenecks on OpenAI's end. If these issues are not promptly addressed, OpenAI risks losing subscribers to competitors that offer better performance and more generous usage policies. OpenAI will need to provide a clear explanation and a plan for resolution to restore user confidence and ensure the competitiveness of its ChatGPT Plus offering for developers.
Beyond the Headlines
The reported performance and quota issues with Codex on ChatGPT Plus underscore a broader industry challenge in scaling AI services while maintaining quality and affordability. As AI tools become more integral to professional workflows, the expectations for reliability and resource availability increase. This incident could prompt a re-evaluation of pricing models and service level agreements for AI subscriptions, pushing providers to be more transparent about performance metrics and usage limitations. It also highlights the competitive pressure within the AI market, where user experience, speed, and cost-effectiveness are key differentiators. The long-term implication could be a shift towards multi-platform AI strategies for developers, reducing reliance on a single provider to mitigate risks associated with performance fluctuations or policy changes.













