On June 3, 2026, three major AI coding assistants rewrote their pricing rules within hours. Cursor lowered its free completion cap, GitHub Copilot confirmed a new agent usage multiplier, and OpenAI Codex switched its free plan to daily credits. The changes arrived without a grace period, according to official changelogs and pricing pages. Developers who relied on free access for daily work found their quotas cut before they could adjust. This was not a staggered rollout. It was a simultaneous shift across the tools that dominate AI coding. The move followed weeks of speculation about rising inference costs and usage-based billing tests across the industry. Free users were hit first, but paid plans also faced new overage fees. The changes reflected a broader pricing push that had been building since May 2026. For more on how these shifts affect developer budgets, see our breakdown of coding tool pricing impact.
Cursor’s free plan fell from 2,000 fast completions per month to 250. GitHub Copilot Free kept its 2,000 completion cap but introduced a 1.6x multiplier on agent mode requests, meaning one agent session could burn 1.6 times the standard quota. OpenAI Codex Free moved from unlimited but throttled access to 50 daily task credits. Those numbers changed the math for students, indie developers, and anyone evaluating whether to pay. Many free-tier users had treated these tools as unlimited utilities, but the new caps converted them into metered products. The reductions hit hardest for long coding sessions, where fast completions and agentic runs are the core value. For a deeper look at shrinking free access, read our free tier limits report.
The timing was not random. OpenAI had signaled Codex changes in its release notes. GitHub posted the Copilot multiplier detail on May 29, 2026. Cursor followed on June 3. All three companies pointed to the cost of agentic inference and the need to protect paid capacity. But the practical result was blunt. Developers who used these tools for free yesterday now had meters. The changes also fueled questions about whether free AI coding tools can survive without ads or usage caps. For a wider view of June pricing shifts, see our free AI pricing changes tracker.
How Do the Top Options Compare?
| Tool | Best For | Free Tier Change | Paid Plan Change | Enforcement Date |
|---|---|---|---|---|
| Cursor | AI-native IDE with fast completions | Monthly fast completions dropped from 2,000 to 250 | Pro $20/mo adds $0.04 per extra fast completion | June 3, 2026 |
| GitHub Copilot | VS Code and GitHub workflows | 2,000 completions kept, 1.6x multiplier on agent mode | Pro $10/mo gets 300 requests, then $0.05 each | June 3, 2026 |
| OpenAI Codex | Agentic coding inside ChatGPT | Unlimited throttled access replaced by 50 daily credits | Plus $20/mo includes 250 credits, then $0.06 each | June 3, 2026 |
Prices and limits reflect public pricing pages as of June 3, 2026. 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. Cursor , AI-native IDE users who need fast completions and deep codebase context
On June 3, 2026, Cursor updated its official pricing page with major free tier reductions. The free plan’s monthly fast completions dropped from 2,000 to 250. Slow requests stayed available but became more heavily rate limited. The Pro plan remained $20 per month, but it added an overage charge of $0.04 for each fast completion beyond the included 500 requests. Business users saw a separate agent credit pool introduced at $30 per 1,000 agentic actions. Cursor’s changelog framed the move as necessary to keep fast inference responsive for paying users, but the cut removed 87.5% of the free fast completion quota overnight.
Free users felt the change immediately. Anyone who burned through 250 fast completions in the first week had to wait until the next monthly reset or upgrade. The slower model remained usable, but it was not a substitute for the fast path. Many developers reported that the reduced free tier pushed them to evaluate Zed and Windsurf, where free tiers were less restrictive at that date. For a comparison of those options, see our Cursor, Windsurf, and Zed free tier analysis.
The overage model also changed paid behavior. A Pro subscriber who regularly used 1,200 fast completions per month would pay an additional $28 on top of the $20 subscription. That was not subtle. It turned the flat Pro plan into a partially metered service. The change brought Cursor closer to the usage-based billing patterns already spreading through the industry, as outlined in our API free tier limits report. Cursor did not offer a grandfather clause for existing free users. The move was part of a broader trend in AI coding tools pricing, which we tracked in the June pricing impact guide.
Key strengths:
- ✅ Fast completions still available on paid Pro plan
- ✅ Monthly price stayed at $20 with no upfront hike
- ✅ Slow requests remained available for free
- ✅ Business plan added predictable agent credit pool
- ❌ Free fast completions cut 87.5% with no grace period
- ❌ Pro overage fees can add $28 or more per month
- ❌ Heavy rate limiting on slow free requests
Who it’s for: Developers who need fast AI completions and are willing to pay for overage, but free-tier users should look elsewhere.
2. GitHub Copilot , Developers inside VS Code and GitHub who need agentic coding workflows
On May 29, 2026, GitHub posted a changelog entry that outlined a new usage multiplier for Copilot. On June 3, 2026, enforcement began. The free plan kept 2,000 code completions per month but introduced a 1.6x multiplier on agent mode requests. That meant a 10-minute agent session that would normally count as one request now consumed 1.6 requests. GitHub Copilot Pro stayed at $10 per month, but it moved to a soft cap of 300 premium requests per month. After that, GitHub charged $0.05 per additional request. Business and Enterprise plans received a pooled agent quota of 10,000 requests per month with $0.08 per extra agentic request.
The multiplier drew immediate backlash. Developers calculated that agent-heavy workflows would burn through free quotas 60% faster than before. A user who used 100 agent requests per day on the free plan would hit the 2,000 completion equivalent in just over three days. The change effectively made Copilot Free less viable for agentic coding, even though GitHub did not reduce the nominal completion cap. Our GitHub Copilot multiplier backlash report details the community reaction.
GitHub framed the multiplier as a fairness measure. The company argued that agent mode requests consumed far more compute than simple completions and that a flat count made free users subsidize heavy agent users. But the result was a hidden cost for developers who had not read the changelog. Many only discovered the multiplier when their quota dropped faster than expected. The shift also aligned with GitHub’s broader move to usage-based billing, which we covered in our June Copilot billing guide. Existing Copilot Individual subscribers were moved to the new model on their next renewal date.
Key strengths:
- ✅ Copilot Free still includes 2,000 base completions
- ✅ Pro plan remains $10 monthly with no price increase
- ✅ Pooled agent quota helps business teams budget
- ✅ Deep VS Code and GitHub integration unchanged
- ❌ 1.6x multiplier cuts effective free quota sharply
- ❌ Overage fees start at $0.05 per premium request after 300
- ❌ Multiplier change was not prominent for existing users
Who it’s for: Developers who use completions more than agent mode and already pay for Copilot Pro or Business.
3. OpenAI Codex , Agentic coding tasks inside ChatGPT and OpenAI API workflows
On June 3, 2026, OpenAI moved Codex to a task credit system across free and paid ChatGPT plans. The free tier shifted from unlimited but throttled access to 50 daily task credits. Each credit covered one Codex agent run up to 10 minutes or 100 tool calls. ChatGPT Plus at $20 per month included 250 monthly task credits, with additional credits priced at $0.06 each. ChatGPT Pro included 1,000 credits per month. The Codex API retained its $0.50 per million input tokens and $3.00 per million output tokens for code models, according to OpenAI’s pricing page.
The free tier change effectively ended unlimited Codex usage for non-paying users. A developer who ran more than 50 agent tasks in a day hit a hard stop, regardless of how light each task was. That was a sharp departure from the earlier throttled model, which allowed long sessions with slower responses. The new system made free Codex a trial product rather than a daily driver. For details on the free tier shift, see our ChatGPT Codex free tier report.
OpenAI framed the change as a way to manage agentic compute costs, which had grown faster than standard chat. The company pointed to the high token consumption of code agents and the need to allocate capacity to paying users. But the move also opened a new revenue line for Codex overages. A ChatGPT Plus user with heavy agent usage could easily add $60 to a $20 monthly bill if they needed 1,000 extra credits. The change fit the wider industry pattern of free tier tightening, covered in our June developer pricing impact guide. 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:
- ✅ Codex free plan still gives 50 daily task credits
- ✅ API token prices did not increase in June 2026
- ✅ ChatGPT Pro includes 1,000 monthly credits
- ✅ Agent runs capped at 10 minutes or 100 tool calls for predictability
- ❌ Free unlimited throttled access ended
- ❌ Hard daily credit cap can interrupt long coding sessions
- ❌ Overage at $0.06 per credit adds up quickly for Plus users
Who it’s for: Developers who want agentic coding with clear credit limits and are already paying for ChatGPT Plus or Pro.
Frequently Asked Questions
What changed for Cursor free users in June 2026?
Cursor cut the free plan monthly fast completions from 2,000 to 250 on June 3, 2026. Slow requests stayed available but became more heavily rate limited. The change removed 87.5 percent of the free fast completion quota overnight.
How does GitHub Copilot's 1.6x multiplier work?
Agent mode requests on Copilot now consume 1.6 times the standard request quota. A 10-minute agent session that once counted as one request counts as 1.6 requests. Enforcement started June 3, 2026 after a May 29 changelog post.
What are OpenAI Codex daily task credits?
Each Codex task credit covers one agent run up to 10 minutes or 100 tool calls. Free ChatGPT users receive 50 credits per day. Once the daily limit is reached, Codex access stops until the next day.
Did paid plans avoid the pricing changes?
No. Cursor Pro added overage fees, GitHub Copilot Pro moved to a 300 request soft cap with per-request charges, and ChatGPT Plus included 250 monthly Codex credits with overage pricing. Paid users gained more capacity than free users, but they lost unlimited or flat usage.
When did the new pricing take effect?
Most changes took effect on June 3, 2026. GitHub posted its multiplier update on May 29, 2026 and began enforcement June 3. Cursor and OpenAI applied their changes immediately on June 3.
Can developers still use free AI coding tools without paying?
Yes, but with much lower limits. Cursor free users get 250 fast completions per month, Copilot Free includes 2,000 completions with the 1.6x agent multiplier, and Codex Free gives 50 daily credits. Long coding sessions on free plans are now impractical for many users.
What Should You Remember?
- Cursor free tier cut 87.5%: Monthly fast completions fell from 2,000 to 250 with no grace period.
- Copilot agent multiplier: Agent mode requests now consume 1.6x standard quota, making free agentic use short lived.
- Codex daily credit cap: Free users got 50 daily task credits, replacing unlimited throttled access.
- Overage fees arrived: Cursor Pro, Copilot Pro, and ChatGPT Plus all added per-request or per-credit overage charges.
- Small teams hit hardest: Indie devs and freelancers lost the most free capacity overnight.
- No grandfathering: Existing free users were moved to new limits on June 3, 2026.
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.
