OpenAI rolls out GPT-5.6 Luna/Sol upgrade with unlimited free ChatGPT chats

OpenAI split ChatGPT's default model by subscription tier this week. Free and Go users now get GPT-5.6 Luna as their default, paired with unlimited text-based chats and a new Think button that gives Luna extra time to reason through harder questions ^[1]. Plus and Pro subscribers move to the more capable GPT-5.6 Sol, which now handles both instant replies and deep reasoning in a single model, plus a new reasoning-effort slider ^[1]. The rollout is staggered - Sol and the slider reached Plus/Pro on August 7 ^[2], with unlimited Luna chats and the Think button following for Free/Go a day later ^[3].

OpenAI frames the update primarily as a reliability fix rather than a raw capability leap. The company reports that responses with at least one factual error dropped 62% for GPT-5.6 Luna and 68% for GPT-5.6 Sol compared with GPT-5.5 Instant, with the biggest gains on financial, medical, and legal questions ^[1][4]. Sol also reportedly posted a perfect 1.000 on a prompt-injection connector evaluation and 0.910 on a search and function-calling evaluation ^[4]- numbers OpenAI is using to argue Sol is safe to hand real tasks to, not just chat with.

This move follows a specific pricing decision: on July 30, 2026, OpenAI cut GPT-5.6 Luna's API price by 80% and Terra's by 20% ^[5], a cut that coverage ties directly to a broader model price war, with Luna positioned as OpenAI's answer to cheaper competitors after Anthropic's Claude Opus 5 and Google's Gemini 3.6 Flash and 3.5 Flash-Lite launches ^[6]. Extending that same economics to the free ChatGPT app lands right as ChatGPT is reported to have crossed roughly 1 billion weekly active users ^[7]. It's also worth noting this is a sequel, not a debut - GPT-5.6 first shipped publicly on July 9, 2026 in three tiers (Sol, Terra, Luna) alongside the ChatGPT Work agent ^[8], so this week's news is really a redistribution of an existing model family down to the free tier rather than a new model launch.

"Unlimited" comes with fine print. OpenAI says unlimited chats remain subject to what it calls abuse guardrails, and hard caps stay in place for file uploads, images, voice, and image generation even on the free tier ^[4]. The company also rolled out new age-based safety guardrails alongside the model update, restricting romantic roleplay, age-restricted challenges, sexual content, eating-disorder and body-image content, and dangerous-activity content for users it detects as being under 18, backed by new under-18 performance evaluations ^[4].

Not everyone reads this as pure generosity. Discussion around the rollout pointed out that GPT-5.6 Luna is a small, low-cost model - essentially Sol with reasoning turned off - not the frontier model powering paid tiers, and that "unlimited" explicitly excludes coding-agent products like Codex. A recurring criticism is that giving Free and the $8-12/month Go tier the identical default model erodes Go's reason to exist, and some observers frame the whole package as an aggressive, possibly unsustainable land-grab timed ahead of a rumored OpenAI IPO rather than a stable pricing decision.