On June 15, 2026, Anthropic ended the flat-rate Claude subscription model that many users had relied on for two years. The company replaced unlimited-style Pro and Max access with a unified credit pool across Claude.ai, Claude Code, and agentic tools. According to Anthropic’s official pricing page, updated the same day, every prompt, file upload, and tool call now drew down a single credit balance. This was not a quiet tweak. It was a structural change that hit paying subscribers first. The overhaul arrived weeks after competitors such as OpenAI and Google adjusted their own pricing, but Anthropic’s move was sharper. See the full credit overhaul for the initial announcement details.
The new pricing affected four main groups: Free users, Claude Pro subscribers, Max subscribers, and Team seats. Claude Pro remained $20 per month but now included 1,500 credits instead of a five-hour rolling limit. Max stayed at $100 per month but included 8,000 credits. Free accounts received 100 credits every five hours, down from the more generous five-hour message reset that had defined the free tier for much of 2025 and early 2026. Team seats cost $30 per user monthly and included 2,500 credits per seat. Those numbers came directly from Anthropic’s pricing page on June 15. See Claude free tier changes 2026 for the free-tier reset mechanics. Existing subscribers were migrated to the new system immediately, with no grandfathering.
Why did Anthropic do this now? The answer was agentic usage. Claude Code, computer use, and third-party agent integrations had driven a surge in token consumption that flat-rate plans could not sustain. Anthropic had already experimented with separate billing for agent workloads in May 2026. The June 15 credit pool formalized that split. Instead of separate limits for chat, coding, and agents, subscribers watched a single balance tick down. Heavier tool calls carried higher multipliers. A single Opus 4.8 agent action could consume 25 credits before token costs. That ended what one Anthropic support document called the agent subsidy. The change was documented on Anthropic’s site on the same day.
The credit overhaul landed inside a broader wave of AI access tightening. Google cut free Gemini Pro API access in June, and OpenAI pushed usage-based billing for Codex. Anthropic’s move aligned with that direction but hit consumers harder because it applied to chat plans, not just APIs or developer tools. Free users were not spared. The five-hour reset that many had learned to navigate gave way to a strict credit allowance. In practice, a heavy Claude Code session could burn a free user’s entire 100 credits in minutes. We covered the parallel free-tier squeeze in AI free tier limits get tougher and the subscription comparisons in AI subscription tiers compared.
How Do the Top Options Compare?
| Plan | Monthly Price | Credit Allowance | Reset Window | Best For |
|---|---|---|---|---|
| Claude Free | $0 | 100 credits | Every 5 hours | Casual chat only |
| Claude Pro | $20 | 1,500 credits | Monthly | Moderate coding and research |
| Claude Max | $100 | 8,000 credits | Monthly | Heavy Claude Code and agents |
| Claude Team | $30 per user | 2,500 credits per seat | Monthly | Small teams with shared billing |
Anthropic applied different credit multipliers by model. Opus 4.8 consumed 2.0 credits per million tokens, Sonnet 4.5 consumed 1.0 credit per million tokens, and Haiku 4.0 consumed 0.4 credits per million tokens. Agentic tool calls added a flat 25-credit charge per action on top of token usage. Prices reflected Anthropic’s official pricing page on June 15, 2026.
1. Claude Free , Casual chat and light document work
Free users kept a $0 plan but lost the generous five-hour reset. On June 15, 2026, Anthropic replaced that rolling window with 100 credits every five hours. Each message, upload, or tool call reduced the balance. The change was announced on Anthropic’s official pricing page. For light chat, the allowance was enough. For Claude Code or longer document analysis, it disappeared quickly. We detailed the earlier free-tier mechanics in Claude free tier changes 2026.
The new credit math made free-tier limits explicit. A standard Sonnet 4.5 message cost around 1 credit. A longer Opus 4.8 message could cost 10 credits or more. File uploads added extra token costs. The old system hid those costs behind a vague five-hour window. The new system made every action visible. Many free users reported hitting the 100-credit cap within the first hour of a coding session.
Anthropic did not offer a way to roll over unused free credits. Each five-hour cycle reset the balance to 100. If a user burned 90 credits in four hours, they waited one hour for the reset. That was better than a daily cap but stricter than the old pain threshold. The move pushed heavier users toward Pro or Max. Anthropic free tier policy console credits covered the policy details.
Key strengths:
- ✅ Zero cost access to Claude models
- ✅ Clear 100-credit allowance replaces vague five-hour limits
- ✅ Five-hour reset keeps some flexibility
- ✅ No credit card required
- ❌ Heavy Claude Code or agent use hits the cap fast
- ❌ No rollover of unused credits
- ❌ File uploads and Opus messages consume credits quickly
Who it’s for: This plan suits users who only need occasional chat or light document summaries, not coding or agent work.
2. Claude Pro , Moderate individual users and part-time coders
Claude Pro stayed at $20 per month, but the value changed substantially. On June 15, 2026, the plan shifted from a five-hour rolling message limit to 1,500 monthly credits. Anthropic’s pricing page listed the change as a simplification, but many users saw it as a cap. Under the old system, a subscriber could send dozens of messages every five hours without tracking a balance. Under the new system, a heavy Opus 4.8 user could exhaust 1,500 credits in a few days.
The credit math made Pro a worse deal for power users. Opus 4.8 consumed 2 credits per million tokens. A single long coding file could cost 100 credits. Claude Code sessions with tool calls added 25 credits per action. That meant a single afternoon of agentic coding could burn 500 to 800 credits. For casual chat, 1,500 credits lasted all month. For serious work, it did not. We compared the new Pro limits in Claude Code limits jump 50 paid vs free 2026.
Anthropic offered no rollover for Pro credits. The balance reset on the billing date. Users who ran out early faced a choice: wait until the next month or upgrade to Max. The company did not sell one-time credit top-ups for consumer plans on launch day. That frustrated users who needed extra capacity mid-month. OpenAI’s ChatGPT Plus, by contrast, kept a capped but higher message limit for $20 monthly. We tracked the subscription comparisons in OpenAI, Anthropic, Google, xAI pricing changes.
Key strengths:
- ✅ Retains $20 monthly price point
- ✅ Clear credit counter replaces ambiguous five-hour window
- ✅ Good for moderate chat and light coding
- ✅ Full Claude Code access included
- ❌ 1,500 credits can vanish fast with Opus or agents
- ❌ No rollover or top-up option at launch
- ❌ Heavy users must upgrade to Max
Who it’s for: Choose Claude Pro if you use Claude most days for chat, writing, or light coding, but not for sustained agent sessions.
3. Claude Max , Heavy Claude Code and agentic workloads
Claude Max remained the premium individual plan at $100 per month. On June 15, 2026, it received 8,000 monthly credits. That sounded generous compared to Pro’s 1,500, but the new multipliers made it less than it appeared. Anthropic confirmed on its pricing page that Opus 4.8 used 2 credits per million tokens. Agentic tool calls added 25 credits each. A heavy Claude Code user running multiple parallel sessions could still burn through 8,000 credits in about two weeks.
The plan was pitched as the replacement for the old Max’s flat-rate unlimited access. Under the previous model, Max subscribers paid $100 for near-unlimited use within a five-hour window. The June 15 change introduced a hard monthly ceiling. Once the balance hit zero, access stopped until the next billing cycle. Anthropic did not offer a pay-as-you-go extension for consumer Max accounts on launch day. The Anthropic ends agent subsidy story covered the policy shift.
For many professionals, Max still offered the best value among Anthropic plans. It included Claude Code, longer context windows, and priority access during peak hours. But the credit system meant users had to budget. Some teams reported that a single developer using Claude Code for six hours daily consumed roughly 350 to 500 credits. That put a monthly ceiling around 16 to 22 full workdays. Google’s comparable Gemini Advanced plan at $25 per month looked cheaper, but model quality and agentic depth differed. See Google AI price cuts for context.
Key strengths:
- ✅ 8,000 monthly credits handle most heavy workloads
- ✅ Includes Claude Code and priority access
- ✅ Lower per-credit effective cost than Pro
- ✅ Same $100 price as before
- ❌ Hard monthly ceiling replaced near-unlimited access
- ❌ Agent multipliers can drain credits quickly
- ❌ No consumer top-up option at launch
Who it’s for: Max suits developers, researchers, and professionals who need Claude Code or agent tools every day and can track a credit budget.
4. Claude Team , Small teams with shared billing and admin controls
Claude Team seats moved to $30 per user per month on June 15, 2026. Each seat included 2,500 monthly credits. Previously, Team plans offered a pooled message quota with less granular tracking. The new system made individual consumption visible. Admins could see which team member burned credits fastest. Anthropic’s official pricing page listed Team pricing alongside Pro and Max. The change followed the same credit pool logic but added a shared billing dashboard.
The $30 per seat price was higher than Pro’s $20, but the per-seat credit allowance was also higher. Teams received 2,500 credits per person, which made sense for collaborative work. However, credit pooling across seats was not automatic. Some teams expected a shared 2,500 times users pool, but Anthropic kept balances per seat on launch day. That meant one heavy user could run out while a light user had surplus. The Anthropic agent billing split story examined how team agent costs changed.
Team plan users also faced the same agentic multipliers. Claude Code tool calls, computer use, and API-like actions consumed credits faster than chat. Anthropic did not offer a discount for bulk seats beyond the standard price. For teams with five users, monthly cost hit $150 for 12,500 credits total. That compared unfavorably to some API-based workflows. But the plan included collaboration features, shared projects, and admin controls. Major AI API pricing model updates covered the API side separately.
Key strengths:
- ✅ Per-seat credit tracking for admins
- ✅ Includes shared projects and collaboration
- ✅ Higher credit allowance per user than Pro
- ✅ Central billing dashboard
- ❌ No automatic credit pooling across seats
- ❌ Agentic multipliers hit team budgets hard
- ❌ Higher per-user cost than Pro
Who it’s for: Choose Team if you need admin controls, shared workspaces, and per-seat usage tracking for a small group.
Frequently Asked Questions
What exactly changed on June 15, 2026?
Anthropic replaced flat-rate Claude Pro and Max access with a unified credit pool. Pro stayed at $20 monthly with 1,500 credits. Max stayed at $100 with 8,000 credits. Free users now get 100 credits every five hours instead of the old message reset. Every chat, upload, and tool call draws from the credit balance.
Do Claude credits roll over month to month?
No. Consumer plan credits reset each billing cycle. Unused Pro or Max credits do not carry forward. Free credits reset every five hours, but unused credits in that window also do not roll over.
How do agentic tool calls consume credits?
Agentic actions cost more than standard chat. Opus 4.8 uses 2 credits per million tokens, Sonnet 4.5 uses 1 credit, and Haiku 4.0 uses 0.4 credits. Each tool call adds a flat 25-credit charge on top of token consumption. This ended the previous agent subsidy.
Can I buy extra credits if I run out?
On June 15, 2026, Anthropic did not offer one-time credit top-ups for consumer plans. Users who ran out had to wait for the next billing cycle or upgrade to a higher plan. Team and API customers had separate billing options.
Did existing subscribers keep their old plans?
No. Anthropic migrated existing subscribers to the new credit system immediately on June 15, 2026. There was no grandfathering. Users logged in to find the credit counter active and their old five-hour windows gone.
Is Claude Free still worth using?
For casual chat, yes. Free users get 100 credits every five hours, which covers light use. For coding or heavy document work, the cap hits fast. Many free users reported burning the allowance in under an hour with Claude Code.
What Should You Remember?
- June 15, 2026: Anthropic ended flat-rate Pro and Max access and introduced a unified credit pool.
- Pro pricing: Pro stayed at $20 monthly but now includes 1,500 credits instead of a five-hour rolling limit.
- Max pricing: Max stayed at $100 monthly but includes 8,000 credits, with a hard monthly ceiling.
- Free tier: Free users now receive 100 credits every five hours, a tighter allowance than the old reset.
- Agent multipliers: Agentic tool calls add a 25-credit charge per action, ending the previous subsidy.
- No rollover: Unused credits do not carry over for consumer plans, forcing careful budgeting.
- Team seats: Team plans cost $30 per user monthly with 2,500 credits per seat and per-seat tracking.
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.
