{"slug": "codex-gpt-5-6-sol-reduced-to-258k-context-window", "title": "Codex GPT 5.6 Sol Reduced to 258K Context Window", "summary": "OpenAI silently reduced the usable context window for its GPT-5.6 Sol model in Codex CLI from 353,400 tokens to 258,400 tokens, a 26.9% cut, despite the model being advertised with a 1.05 million-token context window. The reduction was discovered by a user who reported it as a severe regression, noting that the change was made without notice and undermines trust in the platform for professional development tasks.", "body_md": "-\n[Notifications](/login?return_to=%2Fopenai%2Fcodex)You must be signed in to change notification settings -\n[Fork 14.5k](/login?return_to=%2Fopenai%2Fcodex)\n\n# 🚨 [SEVERE REGRESSION] GPT-5.6 Sol context cut again: 353K → 258K despite advertised 1.05M #32806\n\n[CLIIssues related to the Codex CLI](https://github.com/openai/codex/issues?q=state%3Aopen%20label%3A%22CLI%22)Issues related to the Codex CLI\n\n[bugSomething isn't working](https://github.com/openai/codex/issues?q=state%3Aopen%20label%3A%22bug%22)Something isn't working\n\n[contextIssues related to context management (including compaction)](https://github.com/openai/codex/issues?q=state%3Aopen%20label%3A%22context%22)Issues related to context management (including compaction)\n\n## Description\n\n### What version of Codex CLI is running?\n\n`codex-cli 0.144.3`\n\n### What subscription do you have?\n\nPaid ChatGPT-authenticated Codex subscription. Exact tier is redacted for privacy because the catalog/runtime regression is not account-specific.\n\n### Which model were you using?\n\n`gpt-5.6-sol`\n\n### What platform is your computer?\n\nLinux x86_64. Exact host and kernel details are redacted because they are unrelated to server-delivered model metadata.\n\n### What terminal emulator and version are you using (if applicable)?\n\nCodex desktop/CLI environment. Local application and project identifiers are redacted.\n\n### Codex doctor report\n\nNot included to avoid publishing machine, account, path, repository, and runtime metadata. The relevant sanitized version, model-catalog fields, and runtime values are included below.\n\n### What issue are you seeing?\n\n## Summary\n\nCodex has silently reduced the usable context window for `gpt-5.6-sol`\n\n**again**.\n\nThe previous server-delivered Codex model catalog exposed:\n\n```\n{\n  \"context_window\": 372000,\n  \"max_context_window\": 372000,\n  \"effective_context_window_percent\": 95\n}\n```\n\nThat produced a `353,400`\n\n-token effective runtime window:\n\n```\n372,000 × 95% = 353,400\n```\n\nThe current freshly fetched catalog now exposes:\n\n```\n{\n  \"context_window\": 272000,\n  \"max_context_window\": 272000,\n  \"effective_context_window_percent\": 95\n}\n```\n\nCodex therefore now reports only `258,400`\n\neffective tokens:\n\n```\n272,000 × 95% = 258,400\n```\n\nThis is a further **95,000-token / 26.9% reduction** from the already restricted 353.4K Codex window. It is not UI rounding. Sanitized historical runtime records repeatedly report `model_context_window: 353400`\n\nbefore the rollout and `model_context_window: 258400`\n\nafterward. No local `model_context_window`\n\nor auto-compaction override is configured.\n\nIssue [#31860](https://github.com/openai/codex/issues/31860) already documented that the earlier 353.4K Codex cap was far below the published GPT-5.6 Sol model specification. This report covers the new and materially worse regression from 353.4K to 258.4K.\n\nThe official GPT-5.6 Sol model page still advertises:\n\n- 1,050,000-token context window\n- 922,000 maximum input\n- 128,000 maximum output\n\nSource: [https://developers.openai.com/api/docs/models/gpt-5.6-sol](https://developers.openai.com/api/docs/models/gpt-5.6-sol)\n\nAt the same 95% effective-window policy, the advertised model would provide approximately `997,500`\n\nusable tokens. Codex now exposes only `258,400`\n\n, or **25.9%** of that comparable effective capacity.\n\n## Why this is unacceptable\n\nThis combination is infuriating: OpenAI advertises a 1.05M flagship model, initially restricts it to 353K in Codex, and then silently cuts it again to 258K. Paying users are left with more compactions, lost working context, disrupted long-running tasks, and no advance notice or explanation.\n\nThis feels dismissive of paying users and materially undermines trust in Codex as a professional development tool. Repository-scale analysis, long-running implementation work, multi-agent coordination, large specifications, and tool-heavy sessions all depend on stable context capacity. Context size is not a cosmetic setting that can be silently reduced by another 26.9%.\n\nThe pricing situation makes this even more frustrating. The same official model page states that requests above 272K input are priced at `2×`\n\ninput and `1.5×`\n\noutput for the full request. Codex now caps the raw window at exactly 272K, so users cannot even choose to access the published long-context tier through Codex.\n\nTo be precise: this report concerns the Codex product limit under ChatGPT authentication. The pricing comparison refers to OpenAI's published API long-context policy; it demonstrates that long-context capacity exists and is explicitly priced, while Codex prevents access to it.\n\nMajor competing platforms already offer 1M-class context:\n\n- Google documents Gemini models with 1M or larger context windows:\n[https://ai.google.dev/gemini-api/docs/long-context](https://ai.google.dev/gemini-api/docs/long-context) - xAI documents a 1,000,000-token window for its flagship Grok 4.3:\n[https://docs.x.ai/developers/models/grok-4.3](https://docs.x.ai/developers/models/grok-4.3) - Anthropic documents a 1M context option for Claude Sonnet 4:\n[https://docs.anthropic.com/en/docs/about-claude/pricing](https://docs.anthropic.com/en/docs/about-claude/pricing)\n\nA flagship coding product limited to 258K effective context is no longer competitive, especially when its underlying OpenAI model is publicly specified at 1.05M.\n\n### What steps can reproduce the bug?\n\n- Run Codex CLI\n`0.144.3`\n\nwith the official OpenAI provider and ChatGPT authentication. - Select\n`gpt-5.6-sol`\n\n. - Ensure there is no local\n`model_context_window`\n\n,`model_auto_compact_token_limit`\n\n, provider, or project override. - Fetch or refresh the server-delivered model catalog.\n- Inspect the sanitized\n`gpt-5.6-sol`\n\ncatalog entry. - Observe\n`context_window: 272000`\n\n,`max_context_window: 272000`\n\n, and`effective_context_window_percent: 95`\n\n. - Start or continue a thread and inspect token-usage metadata.\n- Observe\n`model_context_window: 258400`\n\n. - Compare with earlier sanitized GPT-5.6 Sol thread records, which report\n`model_context_window: 353400`\n\n.\n\nThe transition was observed on 2026-07-13. Old and new values briefly overlapped across existing and newly started threads, consistent with a phased or cached server-side catalog rollout.\n\n### What is the expected behavior?\n\n- Codex should expose the full published 1.05M GPT-5.6 Sol context window, with a clear effective input/output reserve.\n- If a product-specific cap is unavoidable, it must be prominently documented before users select the model.\n- Context limits must not be silently reduced for existing or new paid-user workflows.\n- Active threads should retain the context contract under which they were started.\n- Users should be allowed to opt into the published long-context tier when they accept the associated usage or pricing impact.\n- As an immediate stopgap, restore the previous 372K raw / 353.4K effective Codex limit while full 1.05M support is implemented.\n- OpenAI should explain why the context was reduced, whether the change is temporary, which plans/surfaces are affected, and how usage accounting interacts with this cap.\n\n### Additional information\n\nRelated reports:\n\n[[Critical][Codex App] GPT-5.6 Sol is catalog-capped at 372K (353.4K effective) vs the 1.05M model spec #31860](https://github.com/openai/codex/issues/31860)— GPT-5.6 Sol was already capped at 372K raw / 353.4K effective instead of the published 1.05M.[Context window changes from 1M to 258K mid-thread #29039](https://github.com/openai/codex/issues/29039)— a GPT-5.5 thread changed from 1M to 258K mid-thread.[Reopen #19464 and support 1M context #30910](https://github.com/openai/codex/issues/30910)— request to restore 1M context support.\n\nPrivacy note: this report intentionally excludes account identifiers, thread IDs, repository names, local paths, prompts, transcripts, and exact machine metadata. Only model metadata and aggregate context-window observations required to reproduce the regression are included.\n\n## Metadata\n\n## Metadata\n\n### Assignees\n\n### Labels\n\n[CLIIssues related to the Codex CLI](https://github.com/openai/codex/issues?q=state%3Aopen%20label%3A%22CLI%22)Issues related to the Codex CLI\n\n[bugSomething isn't working](https://github.com/openai/codex/issues?q=state%3Aopen%20label%3A%22bug%22)Something isn't working\n\n[contextIssues related to context management (including compaction)](https://github.com/openai/codex/issues?q=state%3Aopen%20label%3A%22context%22)Issues related to context management (including compaction)", "url": "https://wpnews.pro/news/codex-gpt-5-6-sol-reduced-to-258k-context-window", "canonical_source": "https://github.com/openai/codex/issues/32806", "published_at": "2026-07-13 14:50:28+00:00", "updated_at": "2026-07-13 15:05:26.584968+00:00", "lang": "en", "topics": ["large-language-models", "ai-products", "ai-tools"], "entities": ["OpenAI", "Codex CLI", "GPT-5.6 Sol"], "alternates": {"html": "https://wpnews.pro/news/codex-gpt-5-6-sol-reduced-to-258k-context-window", "markdown": "https://wpnews.pro/news/codex-gpt-5-6-sol-reduced-to-258k-context-window.md", "text": "https://wpnews.pro/news/codex-gpt-5-6-sol-reduced-to-258k-context-window.txt", "jsonld": "https://wpnews.pro/news/codex-gpt-5-6-sol-reduced-to-258k-context-window.jsonld"}}