AI Tool Free Tier Limits
What you can actually do without paying. Free-tier limits, model access, training-data policy, and where the paid tier starts — across 14 major AI tools. Sourced directly from each vendor's pricing page.
Verified: 2026-05-06 · Limits taken directly from each vendor's published pricing or feature page. Verified April–May 2026. Vendors change limits without notice — re-check before relying on a number.
| Tool | Category | Free tier summary | Model access | Messages/day | Data trained on? | Paid starts at |
|---|---|---|---|---|---|---|
ChatGPT Free OpenAI | Chat | GPT-4o limited + GPT-4o mini unlimited | GPT-4o (capped, ~5-hour rolling window) · GPT-4o mini unlimited | ~10 GPT-4o messages per 5h window; mini unlimited | ⚠ Yes by default (opt-out in Settings → Data Controls) | ChatGPT Plus $20/mo |
Claude Free Anthropic | Chat | Claude 3.5 Sonnet with daily message cap | Claude 3.5 Sonnet (limited) | ~30–40 messages per 5h window (varies by demand) | ✓ No (Anthropic does not train on user conversations by default) | Claude Pro $20/mo |
Gemini Free Google | Chat | Gemini 1.5 Flash unlimited; 1.5 Pro very limited | Gemini 1.5 Flash unlimited · Gemini 1.5 Pro (very limited) | Flash unlimited; Pro <10/day | ⚠ Yes (default) | Gemini Advanced $20/mo |
Perplexity Free Perplexity | Search | Standard search unlimited; Pro searches 5/day | Default model unlimited; Pro models (Claude/GPT-4o/Sonar Pro) capped at 5 Pro searches per day | Standard unlimited | ⚠ Disclosure varies | Perplexity Pro $20/mo |
GitHub Copilot Free GitHub | Coding | 2,000 code completions/month + 50 chat messages/month (GitHub-verified students get full free) | GPT-4o code model | ≈1.7/day on monthly cap | ⚠ Opt-out available | Copilot Individual $10/mo |
Cursor Free (Hobby) Cursor | Coding | 2,000 completions + 50 slow GPT-4 requests + 200 GPT-3.5 requests / month | GPT-3.5 (200/mo) · GPT-4 slow tier (50/mo) | Capped per-month, not per-day | ⚠ Yes by default; opt-out in Privacy Mode | Cursor Pro $20/mo |
Codeium Free Codeium (Windsurf) | Coding | Unlimited completions + chat for individual developers | Codeium proprietary models (autocomplete + chat) | Unlimited for individual use | ⚠ Yes by default for free tier; opt-out on paid | Teams $12/seat/mo |
Notion AI Free Notion | Productivity | Trial-grade quota (limited responses per workspace) | Notion AI on top of GPT-4o | ≈20–50 responses lifetime per workspace, then upsell | ⚠ Workspace data not used (Enterprise default; Plus opt-out) | Notion AI add-on $10/seat/mo (requires paid Notion seat → effective $20) |
Midjourney Midjourney | Image | No free tier (was discontinued) | — | — | ⚠ Yes (terms grant license) | Basic $10/mo |
DALL-E 3 (via ChatGPT Free) OpenAI | Image | ≈2 generations/day via ChatGPT Free; unlimited via Plus | DALL-E 3 | ~2 image generations on free | ⚠ Yes by default | ChatGPT Plus $20/mo (or API, $0.04/image for 1024x1024 standard) |
Stable Diffusion (Stability AI) Stability AI | Image | Open-weights for self-hosting; cloud credits 25/month free | SD 3 / SDXL / SD 1.5 (self-hosted unlimited) | Unlimited self-hosted; 25 cloud credits/month | ⚠ Open weights — operator controls | Cloud Credit packs from $10 |
Jasper Jasper.ai | Productivity | 7-day free trial; no permanent free tier | GPT-4 / Claude / Jasper proprietary | — | ✓ No | Creator $49/mo |
Copy.ai Copy.ai | Productivity | Free plan with 2,000 words/month | GPT-4 / Claude | Word-cap, not message-cap | ✓ No | Pro $49/mo |
Writesonic Writesonic | Productivity | Free plan with 10,000 words/month | GPT-4o mini / Claude Haiku on free; GPT-4o / Claude Sonnet on paid | Word-cap | ✓ No | Individual $16/mo |
How to read your result
Each row is one AI tool; each column is one dimension of the free tier. The 'free tier summary' column is the headline — what you can actually do without paying. The detailed columns then break that down: which models, how many messages, what files, training-data policy, and where the paid tier starts.
The most important column is 'Data used for training'. Some vendors train on free-tier conversations by default (ChatGPT, Gemini, Cursor) and require an opt-out. Others do not (Claude, Jasper, Copy.ai). For sensitive work, this column matters more than the message cap.
'Paid tier starts at' is the lowest paid tier — not the recommended tier. ChatGPT Plus at $20/mo gives more, but the comparable seat for serious teams is usually $25–30/mo on Team or higher.
When to use this tool
- ›Choosing which AI tool to start with given a workflow — compare the free tiers before committing to a paid one.
- ›Auditing which tools you can drop from a paid stack. If your usage fits inside a free tier on a competitor, the paid seat may be unnecessary.
- ›Privacy-sensitive work — checking 'Data used for training' before pasting anything that shouldn't be in a training corpus.
- ›Comparing real free-tier capability vs. marketing claims. Some 'free' tiers are trial-grade (lifetime quota of ~50 responses); others are genuinely unlimited.
Methodology
All limits are taken directly from each vendor's published pricing or features page (URL listed in the 'Source' column on each row). We do not rely on third-party aggregator data.
Free-tier limits change frequently — vendors raise or lower caps without announcement. The 'verified' date at the top of this page reflects the most recent re-check. We re-verify on a 30-day rolling cadence for AI tools because pricing and limits move faster than fintech rates.
Limits we acknowledge: 'messages per day' is approximate for tools that use rolling-window quotas (e.g., ChatGPT's '5-hour window'). The page reports the typical daily-equivalent cap. For tools with monthly word-counts (Copy.ai, Writesonic) we note 'word-cap' rather than message-cap.
We do not include tools that have no public free tier and no published price (typically enterprise-only). The table is meant to support free-vs-paid evaluation, not enterprise procurement.
Site-wide methodology framework: /methodology/ · Pre-publication standards: /editorial-standards/
FAQ
Why are some 'free tiers' so much better than others?
Two reasons. First, different revenue models — Codeium and Gemini Flash subsidize generous free tiers because they monetize through enterprise/API rather than per-seat consumer subs; Jasper and Copy.ai are pure SaaS plays that need fast paid conversion. Second, model cost — running GPT-4o per request costs OpenAI more than running GPT-4o mini, so the free tier on the cheaper model is more generous. The column to watch is 'model access', not just 'messages per day'.
How do I know if a vendor trains on my conversations?
Check the 'Data used for training' column. As of April 2026: ChatGPT Plus and free trains by default (opt-out in Settings → Data Controls); Claude Free and Pro do not train (Anthropic policy); Gemini trains by default; Cursor free trains by default (privacy mode opt-out on paid); GitHub Copilot allows opt-out. For genuinely sensitive content, treat this as a privacy decision, not a feature comparison.
Why do some message caps say '5-hour window' instead of 'per day'?
Because ChatGPT (and now most large-language-model chat products) use rolling-window quotas to smooth load — instead of '40 messages per day', it's 'X messages per 5 hours'. The practical effect is similar over a full day, but if you hit the cap, you wait the rolling window down rather than until midnight. The table shows the per-day-equivalent.
Can I rely on these limits to plan a workflow?
For a few weeks, yes. Beyond that, no — vendors change limits without notice (raising and lowering both happen). The 'verified' date is a freshness signal, not a guarantee. For workflow-critical limits (e.g., 'we depend on 100 free Claude messages per day'), build in slack and re-check the limit before any major commit.
Why isn't [my favorite tool] in the table?
We focus on the 14 most-searched AI tools as of 2026. Adding a tool requires verifying limits across all eight columns and re-verifying every 30 days, so we cap the table at a maintainable size. If a tool has more search volume than one already listed, we may swap. Suggestions are welcome at the contact email.
How accurate are the 'paid tier starts at' prices?
Accurate to list-price as of the verification date. Most vendors offer 15–20% off for annual billing, which we don't show — actual annual prices are 80–85% of the listed monthly. Sales tax and team/enterprise tier pricing also vary.
Related Tools
Disclaimer: Free-tier limits change frequently and are set entirely by the vendor. The numbers shown reflect the most recent verification date listed above. Always re-check the source pricing page before relying on a specific limit for production work. Zenvestly does not have an affiliate relationship with any vendor in this table.