How much can I use ChatGPT for free?

The Big Change: Free Text Chat Is Now Unlimited

If you've been rationing your ChatGPT messages like they're precious tokens, you can stop. OpenAI no longer enforces rate limits on text-only prompts for Free and Go accounts — users can now "talk with ChatGPT as much as you want," according to OpenAI's announcement. The company's own help documentation confirms it plainly: "Free users have unlimited everyday text chats, subject to abuse-prevention safeguards."

This is a genuine shift in how the free tier works. OpenAI framed the change as "a concrete step toward more abundant intelligence," adding that "access shapes opportunity." For anyone using ChatGPT for daily work — drafting emails, summarizing documents, brainstorming — the practical ceiling on everyday conversations has effectively disappeared.

For context, this replaces a much tighter regime. Earlier reporting described a quota of 10 messages every 5 hours on GPT-5, after which free users were automatically downgraded to a smaller mini model. That structure, documented in a September 2025 analysis, has been superseded by the current unlimited-text policy — so if you see those old numbers quoted elsewhere, treat them as history, not current rules.

A few important caveats keep this from being a total free-for-all:

• Unlimited applies to text-only chats. Separate limits still apply to file uploads, image generation, voice, and data analysis — OpenAI doesn't publish the exact numbers, but shows you when you've hit them and when they reset.
• "Think" mode for harder questions is unlimited when used in text-only chats, but combining it with other features counts against a separate limit.
• Abuse-prevention safeguards still apply, so "unlimited" doesn't mean unmonitored.

The model behind all this chatting is GPT-5.6 Luna, which OpenAI describes as the fastest and lowest-cost model in its family and the new default for Free and Go users. Notably, OpenAI hasn't explained exactly how it removed the text limits — though The Information reported the company was close to a breakthrough on inference costs, which likely made the economics work.

For business users, the takeaway is simple: the free tier is now a legitimate daily driver for text-based work. And if you're building agents on top of these models, free-tier users can use existing GPTs (though not create new ones on personal accounts). When you're ready to extend what your agents can actually do, SkillzBench is building a free directory of vetted agent skills plus a free MCP connector — with security scanning and benchmarking underway ahead of launch. You can join the waitlist today.

Where Limits Still Apply: Tools, Models, and Features

Unlimited text doesn't mean unlimited everything. While ChatGPT's free tier now lets you chat as much as you want, OpenAI's own documentation is clear that separate limits continue to apply to file uploads, image generation, voice, data analysis, and other tools — and those caps can still interrupt your workflow.

Here's the frustrating part: OpenAI doesn't publish the exact numbers for any of these tool quotas. Instead of a public quota table, ChatGPT notifies you in-product when you hit a limit and shows a reset timer, per the free tier FAQ. Anyone quoting you a precise "X images per day" figure is guessing — older reporting described around 2–3 DALL-E images per day on the free plan, but that reflects the previous quota structure before the unlimited-text change.

So what do free users actually get? Quite a lot, as it turns out:

• GPT-5.6 Luna as the default model — the fastest, lowest-cost model in the GPT-5.6 family, now standard for Free and Go accounts (though GPT-5.6 Sol and GPT-6 remain off-limits)
• Unlimited "Think" in text-only chats — combining Think with other tools counts against a separate limit, but Think alone in pure text chats is unrestricted
• Web search, data analysis, image creation, and file uploads (within their separate quotas)
• A library with 500 MB of storage for your saved work
• Use of existing GPTs — but you cannot create or publish new ones, a restriction that applies to Free, Go, Plus, and Pro personal accounts alike

Why do limits exist at all? Running frontier models requires serious hardware, and industry analysis notes that usage caps help balance global demand and prevent GPU overloads that could slow or crash the system. Rate limits also reflect real inference costs and help control abuse — OpenAI frames its removal of text limits as "a concrete step toward more abundant intelligence," suggesting the economics finally worked for text but not yet for compute-heavy tools.

One practical tip: stay signed in. Logged-out users don't get the full free experience, including access to some models. And if you do hit a tool limit, upgrading to Plus, Pro, or Business resets your usage immediately — OpenAI Support will not reset limits for you.

For readers building agents or workflows, the GPT restriction is worth noting: free and personal accounts can use existing GPTs but can't create new ones. If you want to extend what your AI tools can do without paying for creation access, browsing free agent skills in SkillzBench's directory — and connecting them via its free MCP tool once live — is one route to explore. SkillzBench is building exactly that: verified, benchmarked skills you can test for free, with a waitlist open now.

Six Tactics to Maximize Your Free ChatGPT Access

Unlimited text chat sounds like a free-for-all, but free-tier ChatGPT still has invisible tripwires — and knowing where they are is the difference between a productive afternoon and a locked-out afternoon. Here are six tactics that keep you working inside the quota system instead of fighting it.

1. Stay signed in. According to OpenAI's model documentation, logged-out users lose access to some models entirely — GPT-5.6 Sol requires sign-in even for eligible users. If you're working from a shared device or incognito window, you're quietly downgrading yourself before you've typed a single prompt.

2. Keep heavy tasks text-only. Per OpenAI's Free Tier FAQ, free users get unlimited everyday text chats, subject to abuse-prevention safeguards. But the moment you attach a file, generate an image, or run voice mode, you're drawing from separate tool quotas whose exact numbers OpenAI doesn't publish. The workaround is simple: paste data as text, describe images in words, and save tool usage for when it genuinely matters.

3. Use Think freely — but only in text-only chats. Engadget's reporting on OpenAI's policy change notes that Think is unlimited for free users in text-only conversations, yet combining Think with other features counts against a separate limit. Hard reasoning questions belong in plain text chats, full stop.

4. Watch the in-product reset timers. When you do hit a tool limit, ChatGPT notifies you and shows reset timing directly in the interface. Rather than guessing when your image generation or data analysis quota returns, schedule around the timer — run text-based work while the clock counts down.

5. Know that upgrading resets limits immediately. OpenAI's FAQ states plainly that if you reach a rate limit on a Free tier account, your usage resets when you update to a Plus, Pro, or Business plan. No waiting period, no prorated delay — the upgrade clears the wall instantly.

6. Don't ask Support for a manual reset. OpenAI's documentation is explicit: "OpenAI Support does not reset ChatGPT or Codex usage limits." Planning around resets is your only lever, so treat quota awareness as part of your workflow.

If you're stretching free access because you're building agent workflows, one more constraint matters: free-tier users can use existing GPTs but cannot create or publish new ones — GPT creation lives only in Business, Enterprise, and Edu workspaces. That's why many builders extend their agents with instruction packages and connectors instead. SkillzBench is building a free directory of agent skills — security-scanned, benchmarked, and connected via a free MCP tool — so you can test which ones actually lift performance before committing.

The core discipline is simple: keep the heavy lifting in unlimited text, spend scarce tool quota deliberately, and let the reset timers — not frustration — set your pace.

When Free Hits Its Ceiling: Extending Your AI Setup with Agent Skills

Unlimited text chat is generous — until you want to build something. Free users can discover and use existing GPTs, but OpenAI's documentation confirms that creating or publishing new GPTs is reserved for Business, Enterprise, and Edu workspaces. Personal accounts, including Plus and Pro, don't get that option either.

That's a real ceiling for power users and agent builders. If your workflow needs a custom assistant with specialized instructions, the free tier — and even paid personal tiers — leave you looking elsewhere. The good news is that custom GPTs aren't the only way to extend what your AI can do.

Agent skills are the broader answer. These SKILL.md instruction packages work across Claude Code, Codex, Cursor, Gemini CLI, and chat-app agents, teaching them new capabilities without touching your ChatGPT quota. They're free, community-built, and increasingly everywhere on GitHub. The catch: nobody can tell at a glance which ones actually help and which are risky. Published research found that roughly 26% of skills carry at least one vulnerability, so verification matters as much as discovery.

A sensible approach to adopting skills looks like this:

• Discover skills from a deduplicated source rather than raw GitHub search results
• Verify each skill's version is pinned to a specific commit SHA, with license and security checks
• Benchmark it in a sandbox — test the same model with and without the skill to measure real improvement
• Equip only what you need, per agent, so your AI sees a curated toolset rather than everything at once

This is where SkillzBench fits. It's a free, independent directory of agent skills that crawls and dedupes skills from GitHub, validates and security-scans each one, and benchmarks them in a sandbox to publish the measurable Lift each skill provides. A free MCP connector — currently in development, with the waitlist live — will let you connect any agent to the skills you choose, with per-agent loadouts so each agent only sees what you've toggled on. Everything on the platform is free: the directory, benchmarks, and alerts.

Because free ChatGPT now offers unlimited text chats, your budget is better spent on capability, not conversation volume. Skills — proven, verified, and benchmarked — are how you get more from the tools you already use at no cost.

Skills, proven. Browse the free directory and join the SkillzBench waitlist today.