# GPT-5.6 Sol Retune: What ChatGPT’s Update Means for Developers > Source: > Published: 2026-08-09 07:15:45+00:00 On August 6, OpenAI shipped two ChatGPT changes that sound bigger than they are — and one that sounds smaller than it is. GPT-5.6 Sol got a factual accuracy retune for Plus and Pro users. Free users finally get unlimited text chats on Luna, starting this week. But the line that matters most is buried in the announcement: the Sol model powering your API calls, Codex, and ChatGPT Work is completely unchanged. Before you schedule evaluation reruns, read the fine print. ## What Changed in ChatGPT (The Sol Retune) For Plus and Pro subscribers, OpenAI replaced the awkward Instant-versus-Thinking model split with a single GPT-5.6 Sol and a [reasoning slider](https://openai.com/index/improving-gpt-5-6-sol-in-chatgpt/). The slider lets you set effort per message — from fast answers at the low end to extended reasoning at the top — without switching between what felt like two different products. One model, adjustable depth. OpenAI’s headline accuracy claim: the updated Sol produced 68% fewer factual errors than GPT-5.5 Instant in an internal evaluation covering finance, medical, and legal prompts. That number is directionally useful. It is also internally sourced, with no published prompt set, no sample size, and no third-party validation. Take it as “they improved it meaningfully” rather than a benchmark you can reproduce. What you can test is whether it handles the specific fact-checking workflows you actually run in ChatGPT. ## Free Users Get Unlimited Luna (Starting August 10) The other half of the announcement affects everyone who has not paid for ChatGPT. GPT-5.6 Luna replaces GPT-5.5 as the free tier default, and the old 10-messages-per-5-hours cap disappears the week of August 10. Luna is the entry-level model in the GPT-5.6 family — $0.20 per million input tokens at the API, which is 96% cheaper than Sol. It is not a crippled demo. Free users also get a Think button: a per-message toggle that bumps up reasoning effort for harder questions. The limits that remain are file uploads, image generation, and PDF analysis. Text conversations: no cap. For context, [TechCrunch notes](https://techcrunch.com/2026/08/06/openai-brings-unlimited-chatgpt-text-chats-to-free-users/) that Claude and Gemini both already offered unlimited free text chats before this announcement. OpenAI is catching up to its competitors here, not setting a new standard. ## The Trap Developers Need to Avoid Here is the thing that got lost in most coverage. OpenAI said it explicitly: “This version of GPT-5.6 Sol is for everyday chats, so it will only be available in the Chat experience in ChatGPT. The version of GPT-5.6 Sol that powers Work and Codex is not changing.” This is a product polish update, not a model upgrade. The `gpt-5.6-sol` endpoint in the API has not changed. ChatGPT Work has not changed. Codex has not changed. If you run production pipelines on GPT-5.6 Sol through the API and you read “Sol is 68% more accurate now” — your pipelines are unaffected. [The New Stack put it plainly](https://thenewstack.io/gpt-sol-chatgpt-split/): “GPT-5.6 Sol just got better in one place and stayed the same everywhere else.” Do not rerun expensive evaluations based on a ChatGPT interface update. API pricing is also unchanged: Sol at $5/$30, Terra at $2/$12, Luna at $0.20/$1.20 per million input/output tokens. ## What This Move Is Really About ChatGPT is approaching one billion weekly active users, and the free tier is where that number is won or lost. [The Next Web frames it well](https://thenextweb.com/news/chatgpt-free-unlimited-text-chats-gpt-5-6-luna-default): removing the message cap is a direct competitive response. Anthropic and Google have both offered unlimited free text for months. OpenAI needed to close that gap before users treated the free tier limitation as a reason to stay on Claude or Gemini. The Sol consolidation is a separate but equally clear signal: OpenAI is simplifying its product. One model with a reasoning slider is easier to explain, easier to use, and easier to sell. The complexity of Instant versus o-series versus Thinking modes was a product problem. The slider solves it cleanly. ## What You Should Do **API users:** Nothing changes. No action required, no evaluation reruns needed.**ChatGPT Plus/Pro users:** The reasoning slider is a real improvement. Use the high or extra-high setting for fact-sensitive work — legal review, financial analysis, technical documentation checks. Skip the low end for anything where accuracy matters.**Free tier users:** Upgrade your mental model. Luna is a genuine GPT-5.6 model with meaningful reasoning on demand via the Think button. If you were limiting ChatGPT use because of the cap, that constraint is going away this week. The 68% accuracy improvement is worth watching but not worth betting your workflows on yet. When third-party benchmarks reproduce it, revisit. Until then, test it on your actual prompts and decide from there.