On May 20, 2026, OpenAI removed GPT-5.2 from the free ChatGPT tier and capped free access to a smaller GPT-5 Mini model at 10 messages every three hours. Anthropic followed on June 15, 2026, replacing flat free Claude access with a credit pool that gave free users 42 credits every five hours. Google moved on June 9, 2026, cutting Gemini API free quota by 70 percent and pulling Gemini 3 Pro from free consumer accounts. xAI tiered Grok v9 on June 3, 2026, limiting free users to 15 prompts per two hours and removing agentic coding.
Free plans did not disappear. They got thinner. The flagship models that once sat inside free tiers became paid features. Vendors described the shifts as sustainability moves. Users described them as slow paywalls. The result was the largest single-month restriction of free AI access since the release of GPT-4 in 2023.
Affected users included anyone on ChatGPT Free, Claude Free, Gemini Free, and Grok Free. Developers using free API keys also took direct hits. Google’s API change removed Gemini 2.0 Flash from free access and required paid billing for Gemini 3 Pro. Anthropic’s Console free tier no longer provided flat Claude Opus 4.8 access, according to the official Anthropic announcement. OpenAI’s pricing page confirmed that GPT-5.2 required ChatGPT Plus at $20 per month. The pattern repeated across coding tools as well. GitHub Copilot moved to usage-based billing in June 2026, and free coding tiers lost agentic execution. Free tier limits got tougher across the four major providers tracked the coordinated shift.
These were not quiet adjustments. Community forums and developer threads filled with quota complaints within hours. The sharpest reaction came after Anthropic’s credit pool announcement, which ended the agent subsidy that had made Claude useful for free automation tasks. Free users still had a path, but it was narrow.
Why did flagship models go paid in 2026? The cost per million tokens for frontier runs remained high even as model efficiency improved. Stanford HAI’s 2026 AI Index, published at Stanford HAI, reported that inference costs for top models fell more slowly than training costs. Enterprise demand for agentic workloads grew 41 percent year over year. Providers faced a billing mismatch. Free users consumed agentic compute, but revenue came from developers and enterprises. That mismatch pushed vendors toward paid flagships and credit-based rationing. The shift was also competitive. OpenAI, Google, Anthropic, and xAI all moved within a four-week window, forcing users to compare paid plans rather than default to a free flagship.
For users, the immediate change was simple. The best model is no longer free. Free tiers still offer an older model or a smaller distillation, but the flagship sits behind a subscription. Some open-weight models remained available on Hugging Face, and Mistral kept a free Le Chat tier with a mid-size model. The practical advice for cost-sensitive users is to treat 2026 free tiers as sampling tools, not production subscriptions. Developers on free API keys should not build production code against a free flagship assumption. The full run-down is in major AI model tier changes: flagship models now paid, free tiers get lighter.
How Do the Top Options Compare?
| Provider | Flagship Model | Free Tier Change | Paid Entry Price | Effective Date |
|---|---|---|---|---|
| OpenAI | GPT-5.2 | Removed from free; GPT-5 Mini capped at 10 messages per 3 hours | $20 per month Plus | May 20, 2026 |
| Anthropic | Claude Opus 4.8 | Flat access replaced by 42 credits per 5 hours | $20 per month Pro | June 15, 2026 |
| Gemini 3 Pro | Removed from free; API quota cut 70 percent | $19.99 per month AI Pro | June 9, 2026 | |
| xAI | Grok v9 | Capped at 15 prompts per 2 hours; no agentic coding | $8 per month SuperGrok | June 3, 2026 |
Monthly prices in USD. Limits were current at publication and may shift by region or account age.
1. OpenAI ChatGPT Free and Plus , Best for seeing the widest consumer paywall in action
OpenAI removed GPT-5.2 from the free tier on May 20, 2026. The official OpenAI pricing page listed GPT-5.2 as a ChatGPT Plus feature at $20 per month. Free users now get GPT-5 Mini with a 10-message cap every three hours and no priority access. Long context was also restricted. The previous 1 million token context window stayed on paid tiers.
The move ended a brief period where free users could test the same flagship model as Pro subscribers. It also shifted ChatGPT’s free tier toward a lighter assistant. ChatGPT pricing changes 2026 tracked the update as part of a broader consumer tier overhaul. For free users, the change meant that longer documents, code files, and multi-step research tasks became harder to run without interruption.
Paid users on Plus retained the full flagship model, priority queue access, and longer context. The $20 monthly price stayed flat, but the free plan lost its most valuable perk. Users who refused to pay still had access to GPT-5 Mini. The gap between free and paid widened sharply in a single day.
Key strengths:
- ✅ Keeps a functional free assistant with GPT-5 Mini
- ✅ Clear $20 monthly path to GPT-5.2
- ✅ Paid users retain long context and priority compute
- ❌ Free tier lost the flagship model entirely
- ❌ 10 messages every three hours is restrictive
- ❌ No free access to longer context windows
Who it’s for: ChatGPT users who need the newest OpenAI model should budget for Plus.
2. Anthropic Claude Free and Pro , Best example of a credit pool replacing flat free access
Anthropic announced on June 15, 2026, that Claude free and Console users would no longer receive flat access to Claude Opus 4.8. The official Anthropic announcement described a new credit pool. Free users get 42 credits every five hours. Claude Opus 4.8 costs 10 credits per message. That equals roughly four flagship messages per reset. Claude Sonnet costs fewer credits and became the default free model.
The change ended Anthropic’s agent subsidy. Anthropic ends agent subsidy June 15 credit pool replaces flat rate access reported that free agentic runs were the first casualty. Prior to June 15, free users could run multi-step tasks with Opus for hours. After June 15, four Opus messages made agentic loops impractical.
Paid users on Claude Pro at $20 per month receive five times the free credit pool and priority access to Opus 4.8. The price did not rise, but the free experience got measurably smaller. For developers on Console, the change meant that open projects had to attach billing or lose access to Opus-level agent runs.
Key strengths:
- ✅ Free tier still available with Claude Sonnet
- ✅ Credit pooling makes costs transparent
- ✅ Paid Pro keeps priority usage at the same price
- ❌ Free Opus 4.8 is effectively four messages per reset
- ❌ Agentic free runs ended on June 15
- ❌ Credit pool resets every five hours, not daily
Who it’s for: Claude users who want Opus 4.8 must pay for Pro or ration carefully.
3. Google Gemini Free and API , Best for API developers hit by quota cuts
Google shifted Gemini 3 Pro to paid tiers on June 9, 2026. The Google AI changelog confirmed that Gemini 2.0 Flash would no longer be available on free API keys. Free API quota dropped 70 percent across all Gemini models. Free consumer accounts lost Gemini 3 Pro entirely. The free consumer tier now defaults to Gemini 2.5 Flash with a daily prompt limit.
The API change was the most consequential for developers. Google Gemini API free tier tightened: Pro models now paid detailed the new billing thresholds. Developers on free keys had to attach a credit card to continue calling Gemini 2.0 Flash or to move to Gemini 3 Pro. Many small projects that relied on free API access broke on migration day.
Google’s paid AI Pro plan at $19.99 per month restored Gemini 3 Pro and higher API quota. But the free tier, once generous, became a strict on-ramp. The API quota cut hit side projects, educational tools, and lightweight automations hardest. Google framed the move as a capacity decision. Developers framed it as a forced upgrade.
Key strengths:
- ✅ Paid Pro plan under $20 keeps Gemini 3 Pro
- ✅ Clear migration date for API users
- ✅ Free consumer tier still has a mid-size model
- ❌ 70 percent API quota cut hurt free developers
- ❌ Gemini 2.0 Flash API retired with short notice
- ❌ Free users lost flagship access
Who it’s for: Developers and consumers who rely on Gemini should move to Google AI Pro or paid API billing.
4. xAI Grok v9 Free and SuperGrok , Best for a low-cost subscription gate on a flagship model
xAI began gating Grok v9 on June 3, 2026. Free users were capped at 15 prompts every two hours and lost access to the agentic coding mode. SuperGrok, the $8 per month plan, unlocked Grok v9 with a 100-prompt daily cap. The move followed xAI’s attempt to monetize real-time model access after Grok’s free tier drove large consumer usage.
Grok v9 medium free users 2026 covered the tier split. The free tier did not disappear, but it became a preview. xAI also removed free access to Grok Build 01, the agentic coding mode, pushing that workload to paid API customers. The shift followed a familiar 2026 pattern: free users kept a mid-size model, while the newest flagship moved behind a subscription.
At $8 per month, SuperGrok remained the cheapest paid path to a frontier flagship among the four majors. That price point pressured OpenAI and Anthropic, both of which charge $20 for higher limits. Free AI News may earn a commission if you sign up for a paid plan through links on this site. This does not affect our reporting.
Key strengths:
- ✅ SuperGrok at $8 is a low-cost flagship entry
- ✅ Real-time Grok still available to paid users
- ✅ Free preview remains for light use
- ❌ 15 prompts every two hours is tight
- ❌ Agentic coding left the free tier
- ❌ Daily paid cap at 100 prompts may be limiting
Who it’s for: Cost-sensitive users who want a frontier model should consider SuperGrok.
Frequently Asked Questions
Which flagship AI models went paid in 2026?
OpenAI GPT-5.2, Anthropic Claude Opus 4.8, Google Gemini 3 Pro, and xAI Grok v9 all moved behind paid subscriptions between May 20 and June 15, 2026.
What happened to Claude free users on June 15, 2026?
Anthropic replaced flat free access with a credit pool of 42 credits every five hours. Opus 4.8 costs 10 credits per message, so free users got roughly four flagship messages per reset.
Did ChatGPT free tier lose access entirely?
No. Free ChatGPT still works with GPT-5 Mini, but it is capped at 10 messages every three hours and does not include GPT-5.2 or long context access.
Why did Google restrict the Gemini API free tier?
Google cut free API quota by 70 percent on June 9, 2026. The company said the change was a capacity decision driven by paid customers and enterprise demand.
Can I still use any free flagship model in 2026?
Very few. Some open-weight models on Hugging Face and Mistral’s Le Chat free tier remain available, but the four major consumer flagships no longer sit in free tiers.
Will free tiers get weaker again?
Likely yes. Providers face rising inference costs and growing agentic usage. Future free tiers will probably keep only lighter, slower, or older models.
What Should You Remember?
- OpenAI removed GPT-5.2 from free ChatGPT on May 20, 2026.
- Anthropic replaced free Claude Opus 4.8 access with a 42-credit pool on June 15.
- Google cut free Gemini API quota by 70 percent on June 9.
- xAI capped free Grok v9 at 15 prompts per two hours on June 3.
- Paid entry now starts at $8 per month for SuperGrok and $19.99 or $20 for Google and OpenAI.
- Free users must accept smaller, older, or credit-limited models.
- Developers on free API keys should attach billing or migrate before the next quota change.
Free AI News is an independent editorial publication. Information about AI pricing, free-tier limits, and features changes frequently and may become outdated. Always verify current details through the vendor’s official pages. Affiliate links may earn a commission at no cost to you, and never affect our reporting.
