On June 15, 2026, Anthropic quietly replaced the old flat-rate Claude free tier with a five-hour credit pool that assigns a specific cost to every message. The change hit free users first. Anyone using Claude without a paid subscription now sees a credit counter instead of a simple you have reached your limit message. The move followed weeks of limits tightening across major AI providers, documented in Free AI News coverage on AI free tier limits get tougher June 2026. Anthropic published the new structure on its official pricing page at Anthropic.
Free users received 50 credits per five-hour window under the June 15 change. A fast-mode Claude Opus 4.8 message costs 12 credits. Claude Sonnet 4.5 costs 6 credits. Claude Haiku 4.5 costs 2 credits. That means four Opus fast messages, eight Sonnet messages, or 25 Haiku messages before the timer resets. The previous free plan offered 45 Sonnet messages every three hours with no credit math. Free users now have to budget.
The affected group includes all free Claude users in the United States, United Kingdom, European Union, Canada, and other supported markets. Paying Pro and Max subscribers kept larger credit pools. Developers on the free API tier saw separate changes covered in Claude free tier policy console credits. The business context matters. Anthropic ended its agent subsidy on June 15 and moved Claude Code into the same credit pool. That shift was reported in Anthropic ends agent subsidy June 15 credit pool replaces flat rate access.
Why now? Google had already cut Gemini free API access and OpenAI was testing ChatGPT free-tier ads. Anthropic needed to control inference costs while still giving free users a reason to stay. The new free plan is more transparent in some ways and less generous in others. Independent analysis from Stanford HAI noted that AI compute costs for frontier models rose 38 percent in 2025. That made flat-rate free access harder to defend.
How Do the Top Options Compare?
| Free Plan Component | Credit Cost | Uses per 5 Hours | Notes |
|---|---|---|---|
| Claude Opus 4.8 fast mode | 12 credits | 4 messages | Flagship reasoning sample |
| Claude Sonnet 4.5 | 6 credits | 8 messages | Everyday writing and analysis |
| Claude Haiku 4.5 | 2 credits | 25 messages | Quick drafts and simple tasks |
| Claude Code | 15 credits | 3 tasks | Small coding experiments |
| Claude Artifacts | 0 extra credits | Unlimited with message use | Interactive content and code snippets |
Limits apply per five-hour window. No rollover. 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.
1. Claude Opus 4.8 Fast Mode (Free) , Best for testing Anthropic’s flagship reasoning
Anthropic gave free users a limited taste of Claude Opus 4.8 in fast mode starting June 15, 2026. Each fast-mode response consumed 12 credits from the free 50-credit pool. With 50 credits per five-hour window, a free user could send four Opus fast messages before the reset. That was a real change from earlier in 2026, when Opus-class models were locked behind paid plans.
Fast mode was not the full Opus experience. Anthropic described it as a lower-latency, reduced-context version designed for quick answers. The full Opus 4.8 model remained on Pro and Max plans with longer context and higher rate limits. According to Anthropic’s official pricing page at Anthropic, Pro users received 500 credits per five hours. That bought roughly 41 fast Opus messages or 83 Sonnet messages.
The free Opus fast limit was still notable. Free users could test frontier reasoning without paying. But the four-message cap left little room for long conversations. A single follow-up question ate another 12 credits. Heavy users hit the wall quickly.
For most free users, Opus fast mode was a sampler, not a daily driver. It made sense for one-off comparisons or complex math questions. The broader free plan changes were covered in Claude free tier changes 2026 what new 5 hour reset means.
Key strengths:
- ✅ Test Anthropic’s most capable model without paying
- ✅ Fast mode returns answers quickly
- ✅ Credit cost is easy to calculate
- ✅ Reset arrives every five hours, not daily
- ❌ Only four fast Opus messages per reset
- ❌ Follow-up questions quickly drain the credit pool
- ❌ Reduced context compared with full Opus 4.8
Who it’s for: Free users who want to sample Claude Opus 4.8 for a few hard questions each day.
2. Claude Sonnet 4.5 (Free) , Best for everyday free-tier writing and analysis
Claude Sonnet 4.5 became the workhorse of the free plan after the June 15 credit overhaul. Each Sonnet message cost 6 credits. A free user with 50 credits could send eight Sonnet messages per five-hour window. That was a sharp cut from the previous flat-rate allowance, which offered 45 Sonnet messages every three hours before June 15.
The new limit changed how free users worked. Long editing sessions became impossible without waiting for the reset. A typical research request could consume four or five messages. That left three or four messages for follow-up. Some users reported hitting the cap mid-task.
Anthropic pointed to rising inference costs. The June 15 shift also moved agentic features into the credit pool. Free users who ran Claude Code or used Artifacts saw their Sonnet messages shrink faster. The company detailed the new credit structure on its pricing page at Anthropic.
Compared with paid plans, Sonnet access on Pro cost the same 6 credits per message. The difference was volume. Pro users got 500 credits per five hours. That meant 83 Sonnet messages before a reset. Free users got eight. The gap was wide.
For light users, eight messages was enough. For anyone doing real work, it was not. The change fit a broader pattern across the industry. Major providers tightened free tiers throughout June 2026, as reported in Claude resets rate limits May 2026.
Key strengths:
- ✅ Good balance of quality and speed
- ✅ Familiar Sonnet behavior for free users
- ✅ Same credit cost as Pro plan
- ✅ Eight messages works for short tasks
- ❌ Only eight messages per five-hour window
- ❌ Huge gap between free and Pro volume
- ❌ Long conversations require waiting for reset
Who it’s for: Free users who need a capable assistant for a handful of messages before the five-hour reset.
3. Claude Haiku 4.5 (Free) , Best for quick drafts and high-volume light tasks
Claude Haiku 4.5 was the cheapest model in the free credit pool. Each Haiku message cost 2 credits. With 50 credits per five-hour window, free users could send 25 Haiku messages before the timer reset. That made Haiku the only model with enough volume for a sustained working session on the free plan.
Haiku 4.5 was not as capable as Sonnet or Opus. It handled short emails, formatting tasks, and simple summaries well. Complex reasoning or long documents often required a jump to Sonnet. But for repetitive work, 25 messages provided real utility.
The free plan also preserved access to Artifacts with zero additional credit cost. Artifacts generated by Haiku counted only the message cost. So a free user could produce a simple table, a code snippet, or a draft without spending extra credits. Free tier specifics were covered in Claude free plan limits.
The five-hour reset was consistent across all models. If a user burned 50 credits on Haiku in the first hour, they waited four hours for the next 50. There was no rollover. Unused credits disappeared at reset.
For budget-minded free users, Haiku was the practical choice. It could not replace Sonnet for deep analysis. But it stretched the free allowance far enough for daily small tasks.
Key strengths:
- ✅ 25 messages per reset is workable
- ✅ Lowest credit cost on the free plan
- ✅ Fast responses for simple tasks
- ✅ Artifacts do not add credit charges
- ❌ Weaker reasoning than Sonnet or Opus
- ❌ Still limited by the same five-hour window
- ❌ No credit rollover after reset
Who it’s for: Free users who need many short interactions and do not require frontier reasoning.
4. Claude Code Free Tier , Best for small coding experiments with strict caps
Anthropic pulled Claude Code into the same free credit pool on June 15, 2026. Before the change, free users could run Claude Code with a separate flat-rate subsidy. That subsidy ended. The move was part of Anthropic’s agent billing split, covered in Anthropic agent billing split June 2026.
Under the new system, Claude Code tasks consumed 15 credits per run on the free plan. A free user with 50 credits could launch three coding tasks before the reset. That was a dramatic reduction from earlier free access. Developers who used Claude Code for longer sessions had to switch to Pro or Max.
The free tier still allowed Claude Code use, but only for small experiments. A single bug fix might consume one task. Multi-file refactors often required multiple tasks. Three tasks per five hours was not enough for real development work.
Anthropic positioned the change as necessary to stop subsidizing heavy agentic workloads. The corporate move aligned with wider AI pricing shifts. Many coding tools adopted usage-based billing in June 2026, as reported in AI coding tools pricing GitHub Copilot usage based billing June 2026.
For hobbyist coders, the free Claude Code tier was a trial, not a daily tool. Pro and Max plans removed most of the pain by adding hundreds of credits. Free users could still test basic agent workflows before paying.
Key strengths:
- ✅ Free access to agentic coding remains available
- ✅ Clear credit cost per task
- ✅ Same model quality as paid Claude Code
- ✅ Five-hour reset provides a daily allowance
- ❌ Only three coding tasks per reset
- ❌ No more flat-rate free subsidy
- ❌ Unsuitable for multi-file or long-running work
Who it’s for: Developers who want to test Claude Code on small tasks before buying Pro or Max.
5. Claude Artifacts and Free Tools , Best for visual content generation without extra credits
Artifacts remained free to use on the free plan after June 15, 2026. The credit system charged only for model messages. Generating an Artifact from Sonnet or Haiku did not add a separate credit cost. A free user could produce a working HTML snippet, a chart, or a formatted document within an eight-message Sonnet allowance.
The free plan also kept basic file uploads and image analysis at no additional credit charge. Each upload attached to a message spent the standard model credit. Large files or multiple images could slow the response but did not multiply the cost.
The practical limit was tied to the model. Sonnet produced higher quality Artifacts but cost 6 credits per message. Haiku was cheaper but sometimes needed two or three attempts. Free users had to decide whether quality or volume mattered more.
Anthropic did not change the Artifacts interface itself. The feature still let users edit and iterate inside a side panel. The credit counter sat in the corner as a reminder of the new limits.
For students and casual builders, Artifacts gave the free plan genuine utility. It was not enough for production work, but it was better than a plain chatbot.
Key strengths:
- ✅ No extra credit cost for Artifacts
- ✅ Basic file uploads stay free
- ✅ Side panel editing remains available
- ✅ Can use Sonnet or Haiku for output
- ❌ Output quality tied to the model you choose
- ❌ Message cap still limits total Artifacts per window
- ❌ Heavy editing requires more credits
Who it’s for: Free users who want to create simple documents, charts, or code snippets without paying extra.
Frequently Asked Questions
What changed with Claude's free plan on June 15, 2026?
Anthropic replaced flat-rate access with a credit pool that resets every five hours. Free users get 50 credits per window. Opus 4.8 fast mode costs 12 credits, Sonnet 4.5 costs 6, and Haiku 4.5 costs 2.
How many messages can a free user send in five hours?
With 50 credits, a free user can send four Opus fast messages, eight Sonnet messages, or 25 Haiku messages. Mixed usage draws from the same pool.
Does the free plan include Claude Opus 4.8?
Yes. Free users get limited access to Claude Opus 4.8 in fast mode. Each message costs 12 credits, which allows four fast responses per reset.
What happened to Claude Code on the free tier?
Claude Code tasks now consume 15 credits per run. A free user can launch three coding tasks before the five-hour reset. The previous flat-rate free subsidy ended.
Do unused free credits roll over?
No. Credits reset every five hours. Any unused credits are lost when the new window begins.
How does the free plan compare to Claude Pro?
Claude Pro gives 500 credits per five hours at $20 per month. That allows roughly 41 Opus fast messages or 83 Sonnet messages, far above free limits.
What Should You Remember?
- June 15 credit pool replaced Claude’s flat free access with a five-hour reset and per-message costs.
- Four Opus fast messages is the flagship limit on the free plan at 12 credits each.
- Eight Sonnet messages per reset means light use only for free tier writing and analysis.
- 25 Haiku messages is the only high-volume option for free users who need longer sessions.
- Claude Code free tier now allows only three coding tasks per five-hour window after the agent subsidy ended.
- Pro still dominates with 500 credits per reset, far above the free 50-credit allowance.
- No rollover means free credits disappear every five hours if unused.
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.
