--- title: OpenAI admits GPT-5.6 Sol was burning Codex limits way... description: OpenAI's Thomas Sottiaux issued a public apology and global usage reset for ChatGPT Work and Codex subscribers after GPT-5.6 Sol consumed rate limits far... canonical: https://daily.dev/posts/openai-admits-gpt-5-6-sol-was-burning-codex-limits-way-faster-than-intended-v9waqt7o3 twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: OpenAI admits GPT-5.6 Sol was burning Codex limits way faster than intended | daily.dev og:description: OpenAI's Thomas Sottiaux issued a public apology and global usage reset for ChatGPT Work and Codex subscribers after GPT-5.6 Sol consumed rate limits far... og:url: https://daily.dev/posts/openai-admits-gpt-5-6-sol-was-burning-codex-limits-way-faster-than-intended-v9waqt7o3 og:image: https://api.daily.dev/og/posts/V9waqT7O3.png og:image:alt: OpenAI admits GPT-5.6 Sol was burning Codex limits way faster than intended og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # OpenAI admits GPT-5.6 Sol was burning Codex limits way faster than intended **[Trends](https://daily.dev/sources/trends)** · 2 min read · 5 upvotes · 0 comments ## Summary OpenAI's Thomas Sottiaux issued a public apology and global usage reset for ChatGPT Work and Codex subscribers after GPT-5.6 Sol consumed rate limits far faster than anticipated. Sol's architecture is more aggressive than prior models — it runs longer, makes more tool calls, coordinates subagents, and fires parallel tool calls in code mode — causing power users to exhaust weekly limits in hours. A temporary fix delivers an 18% improvement in typical usage duration, but the 5-hour daily cap returns soon. Heavy users remain frustrated: on the $200/month plan, running 1-2 agents at normal speed for 12-16 hours still hits walls, whereas GPT-5.5 barely dented the same allocation. A secondary complaint is the lack of cost transparency — Codex uses a Credits system with no per-thread breakdown or real-time burn rate, just a percentage bar, unlike Claude's /usage command. OpenAI acknowledged that capability and efficiency don't always improve together, and some issues only surface at real-world scale. ## Content OpenAI's Thomas Sottiaux just issued a public mea culpa and a global usage reset for all ChatGPT Work and Codex subscribers. The short version: GPT-5.6 Sol was eating through rate limits much faster than OpenAI expected, and they're only now getting a handle on why. The explanation is actually pretty interesting. Sol is more aggressive than previous models in ways that compound: it runs longer, makes more tool calls, coordinates subagents, and in "code mode" fires tool calls in parallel rather than waiting. All of that is genuinely useful behavior. It's also why some power users were burning through their weekly limits in a couple of hours. Sottiaux admitted the team was focused on median usage before launch and "missed some cases where the long tail could use significantly more." The 5-hour daily cap was temporarily removed while they investigated; it comes back tomorrow. The fix lands as an 18% improvement in typical usage duration, with some users seeing bigger gains. That's not nothing, but it's also not the structural fix that heavy users were hoping for. User @lucastohdev laid out the frustration in detail: on the $200/month plan, running fewer agents at lower thinking settings and slower speed than before, limits still evaporate. He's running 1-2 High agents at normal speed for 12-16 hours a day and still hitting walls. Under GPT-5.5 on the same plan, he could barely spend his allocation. His read: the periodic global resets and the suspended 5-hour cap are just stopgaps, and once both revert, sustained usage on current plans becomes basically impossible. He also has a point about transparency. Codex runs on a "Credits" system where 1 credit equals $0.04 in API equivalent, but there's no per-thread cost breakdown, no real-time burn rate, and no way to see what a task actually cost. Claude has `/usage`. Codex has a percentage bar. For a product aimed at power users running complex multi-agent workflows, that's a pretty significant gap. Sottiaux's tone was unusually candid: "capability and efficiency do not always improve at the same pace, and some issues only become clear once people are using the model at real-world scale." Fair enough. But the 18% improvement and a reset don't change the underlying math if Sol's architecture is just fundamentally more token-hungry than what the subscription tiers were priced around. --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#openai](https://daily.dev/tags/openai), [#gpt](https://daily.dev/tags/gpt), [#codex](https://daily.dev/tags/codex) [View this post on daily.dev](https://daily.dev/posts/openai-admits-gpt-5-6-sol-was-burning-codex-limits-way-faster-than-intended-v9waqt7o3) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"OpenAI admits GPT-5.6 Sol was burning Codex limits way faster than intended","url":"https://daily.dev/posts/openai-admits-gpt-5-6-sol-was-burning-codex-limits-way-faster-than-intended-v9waqt7o3","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/openai-admits-gpt-5-6-sol-was-burning-codex-limits-way-faster-than-intended-v9waqt7o3"},"datePublished":"2026-07-29T07:34:59.872Z","dateModified":"2026-07-29T07:35:51.856Z","description":"OpenAI's Thomas Sottiaux issued a public apology and global usage reset for ChatGPT Work and Codex subscribers after GPT-5.6 Sol consumed rate limits far...","image":"https://pbs.twimg.com/media/HOYPQUJbkAAa0ly.png","thumbnailUrl":"https://pbs.twimg.com/media/HOYPQUJbkAAa0ly.png","isAccessibleForFree":true,"articleSection":"Trends","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Trends","logo":"https://media.daily.dev/image/upload/s--ZfSp3asX--/f_auto,q_auto/v1780996004/logos/trends?_a=BAMAMiWQ0","url":"https://daily.dev/sources/trends"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/openai-admits-gpt-5-6-sol-was-burning-codex-limits-way-faster-than-intended-v9waqt7o3","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":5},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-agents,openai,gpt,codex","timeRequired":"PT2M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Trends","item":"https://daily.dev/sources/trends"},{"@type":"ListItem","position":3,"name":"OpenAI admits GPT-5.6 Sol was burning Codex limits way faster than intended"}]} ```