On June 15, 2026, Anthropic quietly replaced the Claude free tier’s familiar 5-hour reset window with a monthly console credit meter. Free users who logged into claude.ai or the Anthropic Console saw a credit balance instead of the old countdown timers. The first-party change appeared on Anthropic’s official pricing page and console documentation. It applied globally to free Claude.ai accounts and free API keys. The move ended the era of simply waiting 5 hours to resume prompting. Now, every Haiku, Sonnet, and Opus request drew down a fixed monthly allowance. Users had to track credits like a prepaid phone plan.
The old system rewarded patience. A free user could hit a Sonnet limit, wait until the reset timer expired, and continue chatting. Starting June 15, 2026, patience alone no longer restored access. Anthropic granted 120 free credits per calendar month. Those credits reset on the first day of each month, not every 5 hours. The policy change hit casual users who relied on the free tier for daily drafting and coding help. It also affected free Claude API developers who tested prompts through the console. Anthropic framed the shift as a way to make usage costs transparent and stop resource abuse. The company also removed the separate agent subsidy on the same date, folding agent runs into the same credit pool.
The competitive context mattered. OpenAI, Google, and Anthropic all tightened free-tier limits through May and June 2026. Google cut Gemini API free access on June 3, 2026, and OpenAI pushed some Codex usage behind paid plans. Anthropic’s console credit system arrived as AI free-tier limits got tougher across the industry. Unlike the 5-hour reset, credits gave users one hard number to manage. But it also meant a free user could burn a month’s allowance in a single long coding session. The change was not a price cut. It was a rationing system. For developers, the credit meter exposed exactly how expensive each model interaction really was.
Anthropic’s official pricing page listed the new credit costs side by side with legacy limits. The console displayed a running ledger of deductions. Free users could see that Opus 4.8 cost 10 credits per response, Sonnet 4.5 cost 2, and Haiku 3.5 cost 1. Agentic tasks under Claude Code cost 20 credits per autonomous run, the same jobs that previously had a separate flat subsidy. By making credits visible, Anthropic shifted the mental model from time to tokens. Users now had to budget. That was the point.
How Do the Top Options Compare?
| Plan | Monthly Cost | Credit Allowance | Typical Request Cost | Reset Window |
|---|---|---|---|---|
| Free Console Credits | $0 | 120 credits | Haiku 1, Sonnet 2, Opus 10 | Monthly |
| Claude Pro | $20 | 600 credits | Same plus priority | Monthly |
| Claude Max | $100 | 2,400 credits | Same plus longer context | Monthly |
| Anthropic API Pay-as-you-go | Variable | Billed per credit | Credits priced per token | Never expires with top-up |
Credit costs shown reflect the June 15, 2026 Anthropic console documentation and may vary by region. 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. Anthropic Free Tier Console Credits , Casual users who want free access with predictable monthly limits
Anthropic retired the 5-hour reset on June 15, 2026 and gave free accounts 120 console credits each month. The Anthropic Console displayed the balance in the top-right corner next to the user avatar. Each prompt deducted credits based on the model and output length. A single Haiku 3.5 response cost 1 credit, a Sonnet 4.5 response cost 2, and an Opus 4.8 response cost 10. Long outputs consumed multiple credits, so a verbose Opus answer could wipe 30 credits in one turn. The free tier still included access to Claude Haiku and Sonnet models, but no longer offered unlimited within a 5-hour window. Users who hit zero saw a ‘credit exhausted’ screen with options to upgrade or wait until the first of the next month. Anthropic said the credit system was fairer than time-based resets because it showed exact costs. Critics argued it pushed heavy free users toward paid plans faster than the old model. The change matched a broader industry move toward usage-based free tiers documented in AI API free tier limits.
Key strengths:
- ✅ Full free access without a credit card
- ✅ Transparent credit ledger shows each model’s cost
- ✅ No hourly timer to monitor
- ✅ Predictable monthly allowance for light use
- ❌ 120 credits can vanish after a few Opus requests
- ❌ No rollover of unused credits
- ❌ Agent runs cost 20 credits each, limiting Claude Code tests
Who it’s for: Casual users who want occasional Claude access and are willing to budget a small monthly credit pool.
2. Claude Pro , Regular users who need more monthly credits and priority access
The $20/month Claude Pro plan became the first paid escape hatch from the free credit meter. Pro subscribers received 600 monthly credits, exactly five times the free allowance. The plan retained priority access during peak hours and allowed longer context windows. Anthropic updated its pricing page on June 15, 2026 to list Pro as the default upgrade path for users who hit the free 120-credit wall. Pro users could still run out, but the larger pool meant around 60 Opus responses or 300 Sonnet responses per month. Pro did not remove credit metering entirely. It just raised the ceiling. Anthropic also introduced a credit top-up option for Pro users at $5 for 150 extra credits. That price translated to roughly 3.3 cents per credit. The top-up market replaced the old ‘wait 5 hours’ recovery mechanic. Industry watchers compared the shift to how Google restructured Gemini free tier limits in June 2026. For Anthropic, Pro became a more important revenue buffer as free users were nudged upward.
Key strengths:
- ✅ Five times the free credit allowance
- ✅ Priority access during high-demand periods
- ✅ Optional top-up credits for overage
- ✅ Longer context windows than free tier
- ❌ Still a hard credit cap each month
- ❌ $5 per 150 credits adds up quickly for Opus use
- ❌ No unlimited plan under $100
Who it’s for: Users who use Claude daily and want more runway without jumping to the $100 Max plan.
3. Claude Max , Power users and professionals running Opus or agentic workloads
Claude Max at $100 per month became the heaviest consumer plan with 2,400 monthly credits. That allocation was twenty times the free tier and four times Pro. Anthropic marketed Max for users who needed long Opus conversations, large document analysis, and frequent Claude Code agent runs. The Anthropic console credit overhaul showed Max users consumed the same per-request credit rates but had enough headroom to avoid checking the meter. Autonomous agent runs cost 20 credits, so Max users could run 120 agent tasks before hitting the cap. Max also included early access to new models and the highest priority tier. But the plan was not unlimited. A heavy Opus user generating 240 verbose responses per month would exhaust the 2,400 credits at 10 credits each. Anthropic offered additional top-ups at $5 for 150 credits, same as Pro. Some developers said the $100 price felt high compared to OpenAI and Google subscription tiers. Still, Max was the only consumer plan that made Opus-heavy use remotely practical.
Key strengths:
- ✅ 20x the free credit allowance
- ✅ Practical for daily Opus and agent workloads
- ✅ Early access to new Claude models
- ✅ Highest priority access
- ❌ $100 monthly cost is steep for individual users
- ❌ Agent runs still consume 20 credits each
- ❌ No true unlimited consumer plan
Who it’s for: Professionals and developers who need sustained Opus or Claude Code use without constant credit anxiety.
4. Anthropic API Pay-as-you-go , Developers who want metered API access with no monthly subscription
Developers who outgrew the free console credits moved to the Anthropic API pay-as-you-go tier. The API used the same credit pricing as the console but allowed users to pre-purchase credit packs. Anthropic charged $5 for 150 API credits following the June 15, 2026 update. Unlike the free monthly grant, paid API credits did not expire at the end of the month. They remained on the account until spent. This made the API tier attractive for developers who wanted to batch test prompts without worrying about a monthly reset. The pay-as-you-go model also exposed the true cost of heavy usage. A developer running 100 Opus requests per day would burn 1,000 credits daily, or $33.33 at the $5 per 150 credit rate. Anthropic published a credit calculator in the console to estimate monthly burn. The company’s agent billing split moved autonomous agent runs into the same metered system. Developers who previously relied on free Claude Code access found the API path more expensive but more predictable.
Key strengths:
- ✅ Paid API credits never expire
- ✅ No monthly subscription required
- ✅ Same credit rates as console
- ✅ Works for batch testing and production workloads
- ❌ Costs scale quickly with Opus and agent runs
- ❌ No free monthly grant after initial credits
- ❌ Requires manual top-up or billing setup
Who it’s for: Developers and startups that need flexible, prepaid API access without a flat monthly fee.
Frequently Asked Questions
What changed in Anthropic's free tier on June 15, 2026?
Anthropic replaced the old 5-hour reset windows with a monthly allowance of 120 console credits. Free users now see a credit balance instead of countdown timers. Each request deducts credits based on the model used. The old wait-5-hours recovery mechanic is gone.
How many free credits do I get each month?
Free-tier users receive 120 credits per calendar month. Credits reset on the first day of each month. Unused free credits do not roll over. Paid users can buy additional credits or upgrade to Pro or Max for larger monthly pools.
How much do different Claude models cost in credits?
Haiku 3.5 requests cost 1 credit each. Sonnet 4.5 requests cost 2 credits each. Opus 4.8 requests cost 10 credits each. Longer outputs can consume multiple credits, so a verbose Opus answer may cost more than 10.
Do agent tasks and Claude Code runs cost extra?
Yes. Autonomous agent runs through Claude Code cost 20 credits per run under the June 15, 2026 policy. Anthropic folded the previous separate agent subsidy into the same credit pool. This made free agent testing much more expensive.
Can I buy more credits if I run out?
Paid Pro and Max users can buy top-up packs at $5 for 150 credits. API users can pre-purchase credit packs at the same rate. Free users cannot buy one-off credits; they must wait for the monthly reset or upgrade to a paid plan.
Does the free tier still include Claude Opus?
Yes, free users can access Opus 4.8 but each request costs 10 credits. With only 120 free credits per month, a free user can make roughly 12 Opus requests before hitting the cap. That is much tighter than the old 5-hour reset model.
What Should You Remember?
- Monthly credit cap: Free users got 120 credits per month starting June 15, 2026.
- Per-request pricing: Haiku cost 1 credit, Sonnet 2, Opus 10.
- Agent runs: Claude Code tasks now cost 20 credits each, ending the flat subsidy.
- No rollover: Unused free credits disappeared at month end.
- Top-up path: Paid plans offered 150 credits for $5.
- Industry shift: Anthropic joined Google and OpenAI in tightening free-tier access.
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.
