[
    {
        "role": "user",
        "content": "How%20do%20I%20select%20a%20different%20model%20of%20chatgpt"
    },
    {
        "role": "assistant",
        "content": "Here\u2019s how to select a different model in ChatGPT, step by step\u2014including what to expect, and some common pitfalls to watch out for.\n\n---\n\n## \ud83d\udee0 What \u201cmodel\u201d means in ChatGPT\n\n- ChatGPT uses different versions of underlying AI models (e.g. **GPT-3.5**, **GPT-4**, **GPT-4o**, **GPT-5.5 Instant**, etc.), each with trade-offs in speed, cost, reasoning power, memory, or other features. ([digitaltrends.com](https://www.digitaltrends.com/computing/openai-explains-how-to-use-different-chatgpt-models/?utm_source=openai))  \n- Different subscription plans usually grant access to different models or \u201cflavors.\u201d For example, free accounts might get GPT-3.5 or \u201cInstant,\u201d while paid users can access more advanced models like GPT-4 or Pro/Thinking/Reasoning versions. ([digitaltrends.com](https://www.digitaltrends.com/computing/openai-explains-how-to-use-different-chatgpt-models/?utm_source=openai))  \n\n---\n\n## \u2699 How to switch models in ChatGPT\n\nHere\u2019s the usual workflow, though it may vary slightly depending on your platform (web, mobile, desktop) or subscription tier.\n\n1. **Start a new chat**  \n   You cannot normally change models *in the middle* of an existing conversation. If you want to switch to a different model, begin a fresh chat. ([guides.ai](https://guides.ai/chatgpt-models-explained/?utm_source=openai))\n\n2. **Locate the model selector**  \n   Once you're in a new chat, look for a selector bar usually located either:\n   - At the top of the chat window\n   - In the header of the new chat page\n   - Sometimes in a dropdown menu or settings gear nearby\n\n   This selector might display labels like *GPT-3.5*, *GPT-4*, *Instant*, *Thinking*, *Pro*, etc. ([guides.ai](https://guides.ai/chatgpt-models-explained/?utm_source=openai))\n\n3. **Click or tap to open model options**  \n   Click the current model (often labeled \u201cdefault\u201d or \u201cAuto-select\u201d) to open a menu of available models. ([techradar.com](https://www.techradar.com/ai-platforms-assistants/chatgpt/chatgpt-might-not-be-using-the-model-you-think-and-its-also-hiding-others-in-settings?utm_source=openai))\n\n4. **Choose the model you want**  \n   Select from the available options. Depending on your subscription, you might see more models to choose from. Free users usually have fewer options. ([digitaltrends.com](https://www.digitaltrends.com/computing/openai-explains-how-to-use-different-chatgpt-models/?utm_source=openai))\n\n---\n\n## \ud83d\udd0d New \u201cInstant / Thinking / Pro / Auto-Select\u201d modes\n\nOpenAI has added a simpler way to pick how responses are generated:\n\n- **Instant** = fast responses (less processing, sometimes less depth)  \n- **Thinking** = more in-depth responses  \n- **Pro** = highest detail / reasoning (if available to your plan)  \n- **Auto-Select** = ChatGPT decides which model version to use based on your prompt complexity ([digitaltrends.com](https://www.digitaltrends.com/computing/openai-explains-how-to-use-different-chatgpt-models/?utm_source=openai))\n\nYou\u2019ll usually find these options in a \u201cConfigure\u201d or \u201cSettings\u201d menu near the model selector. ([techradar.com](https://www.techradar.com/ai-platforms-assistants/chatgpt/chatgpt-just-made-it-easier-to-pick-the-right-model-just-like-gemini-does-heres-when-to-use-instant-thinking-or-pro?utm_source=openai))\n\n---\n\n## \u26a0 Things to watch out for\n\n- **Availability depends on your subscription plan**  \n  If you don't have a paid account (Plus, Pro, etc.), you might only see lower-capability models. Some advanced ones are locked behind paywalls. ([guides.ai](https://guides.ai/chatgpt-models-explained/?utm_source=openai))\n\n- **\u201cAuto-select\u201d might override your choice in some cases**  \n  If Auto-Select is on, the system may pick which model handles your conversation without you explicitly choosing. ([techradar.com](https://www.techradar.com/ai-platforms-assistants/chatgpt/chatgpt-might-not-be-using-the-model-you-think-and-its-also-hiding-others-in-settings?utm_source=openai))\n\n- **Model might change mid-conversation if limits are hit**  \n  If you exceed usage quotas of a high-tier model (like GPT-4), ChatGPT may switch to a lower tier (GPT-3.5) automatically. ([reddit.com](https://www.reddit.com/r/ChatGPT/comments/18s65i4?utm_source=openai))\n\n- **New UI changes**  \n  OpenAI updates its interface occasionally. Some users report the model-picker being hidden, moved, or simplified. ([community.openai.com](https://community.openai.com/t/how-to-change-model-within-a-project-folder/1055089/31?utm_source=openai))\n\n---\n\nIf you tell me which platform you're using (web, mobile app, desktop) and which plan (free, Plus, Pro, etc.), I can give you more precise instructions tailored for your situation."
    },
    {
        "role": "user",
        "content": "In%20the%20middle%20of%20a%20codex%20project%2C%20it%20says%20my%20selected%20model%20is%20at%20capacity%2C%20what%20does%20that%20mean%2C%20I%20am%20on%20paid%20plan.%20How%20do%20I%20keep%20working%20on%20it%3F%20How%20long%20will%20it%20be%20at%20capacity%3F"
    },
    {
        "role": "assistant",
        "content": "Here\u2019s a breakdown of what \u201cSelected model is at capacity\u201d means, especially for Codex\u2014and what you can do when it happens.\n\n---\n\n## What \u201cmodel is at capacity\u201d means\n\nWhen you see the message *\u201cSelected model is at capacity\u201d*, it doesn\u2019t mean there\u2019s something wrong with **your account** necessarily. It means that *the particular model instance you\u2019re trying to use* is temporarily overloaded. OpenAI has finite compute and memory resources, and when too many users are requesting that model, it hits a limit. ([ccnavx.com](https://ccnavx.com/errors/codex-selected-model-is-at-capacity?utm_source=openai))\n\nSome additional nuance:\n\n- It\u2019s **not** the same as exhausting your usage quota (like hours or tokens) under your plan. Others have reported this error even when their usage\u2014and limits\u2014were well under their plan\u2019s cap. ([reddit.com](https://www.reddit.com/r/codex/comments/1usvz46/selected_model_is_at_capacity_despite_83_5hour/?utm_source=openai))  \n- It's often due to peak demand, routing constraints on the backend, or capacity limits for that **specific model architecture / tier / reasoning level**. ([ccnavx.com](https://ccnavx.com/errors/codex-selected-model-is-at-capacity?utm_source=openai))\n\n---\n\n## Why it might hit you even if you're on a paid plan\n\nBeing on a paid plan (Plus, Pro, Enterprise, etc.) comes with **better priority**, but that doesn\u2019t guarantee 100% uptime under all conditions\u2014especially with high-demand models. Paid plans help, but model capacity is a separate constraint. ([techradar.com](https://www.techradar.com/pro/openai-will-now-let-your-company-pay-to-ensure-guaranteed-capacity-for-up-to-three-years?utm_source=openai))\n\nOpenAI introduces features like **Priority Processing**, **Guaranteed Capacity**, and **Scale Tier** to help larger/enterprise users minimize interruptions. But you often need specific plan upgrades or commitments (financial or contractual) to access them. ([openai.com](https://openai.com/business/guaranteed-capacity/?utm_source=openai))\n\n---\n\n## How long the \u201ccapacity\u201d issue lasts\n\nThere's no fixed duration. Some influencing factors:\n\n- **Peak load times**: During high traffic, models are more likely to be \u201cfull.\u201d\n- **Backend scaling**: OpenAI regularly adds capacity, spin up additional model replicas, or adds new routes. These operations reduce downtime but aren't instant. ([status.openai.com](https://status.openai.com/incidents/0pqnp2gq?utm_source=openai))  \n- **User plan and region**: Your locality or service tier might affect how quickly capacity is restored.\n\nUsually it\u2019s minutes to at most a few hours\u2014but in rare cases, major incidents can last longer. The good news: OpenAI has publicly reported incidents like this (\u201cCodex 'Selected Model is at Capacity' Error\u201d) which have been resolved. ([status.openai.com](https://status.openai.com/incidents/0pqnp2gq?utm_source=openai))\n\n---\n\n## What you can do to keep working\n\nHere are several strategies to keep working when you hit that error:\n\n1. **Switch to a different model or reasoning tier**  \n   If you\u2019re using \u201cHigh\u201d or \u201cX-High\u201d mode, try stepping down to \u201cMedium\u201d or \u201cHigh\u201d (depending on what\u2019s available). This often works because those less-demanded instances may have free capacity. Reddit users report this helps. ([reddit.com](https://www.reddit.com/r/codex/comments/1usvz46/selected_model_is_at_capacity_despite_83_5hour/?utm_source=openai))\n\n2. **Try later**  \n   Wait 5-30 minutes and try again. Capacity often frees up relatively quickly after peak usage dips.\n\n3. **Use acceleration or priority access if available**  \n   Some plans support \u201cPriority Processing\u201d or \u201cGuaranteed Capacity\u201d add-ons. These give you preferential access to overloaded model instances. ([openai.com](https://openai.com/api-scale-tier/?utm_source=openai))\n\n4. **Check your usage limits**  \n   Make sure your usage for that session (hours, tokens, etc.) hasn\u2019t triggered some threshold. Even though usually the \u201ccapacity\u201d message is unrelated to your personal quota, it's good to confirm you\u2019re not hitting any plan-based limits. ([help.openai.com](https://help.openai.com/en/articles/11369540-using-codex-with-your-chatgpt-plan%25252525252525252525252525252525252525252525252525252525252525252525252525252525252525252525252525252525252525252525252525252525252525252525252525252525252525252525252525252525252525252525252528.csv?utm_source=openai))\n\n5. **Switch regions if applicable**  \n   Some users report switching server regions (if supported by your plan) to access less busy server capacity, though this depends on availability.\n\n6. **Reach out to OpenAI support**  \n   If this issue happens frequently and disrupts your workflow, definitely contact support. Ask if there\u2019s anything plan-specific you can do (e.g. upgrade, add credits, or get priority access). Sometimes if there\u2019s a known backend issue, they can provide estimates for fixes. Reddit threads suggest many users have open support cases over this. ([reddit.com](https://www.reddit.com/r/OpenaiCodex/comments/1v9z12i/gpt56_sol_repeatedly_returns_selected_model_is_at/?utm_source=openai))\n\n---\n\nIf you like, I can check real-time status (for your region or plan) to see whether the model's capacity issue has been officially logged or is being mitigated, so you know how long it might last. Do you want me to pull that up?"
    }
]