On June 16, 2026, developers opened their dashboards and found a changed pricing map. GitHub Copilot no longer offered a flat $10 monthly Pro plan with unlimited completions. The new model charged per flow credit, with agent mode consuming 2.5 credits per request. OpenAI Codex had cut its free tier to 50 agent turns per day on June 2. Anthropic replaced Claude Code’s flat rate with a shared credit pool on June 15. GitHub’s official changelog confirmed the switch. The moves hit individual developers, open-source maintainers, and small teams hardest.
The pricing shifts were not isolated. Cursor and Windsurf tightened free plans on June 9. Cursor limited AI chat to 20 messages per day and removed Claude Sonnet 4.5 from the free plan. Windsurf cut free agent runs in half and moved to a 5-hour reset. Gemini Code Assist shut down its free tier entirely on June 5. Google pushed users to a $22.80 monthly Standard plan. These changes came after months of AI providers warning that agentic coding costs were 10 to 30 times higher than simple autocomplete.
Why now? The business context was simple. Anthropic’s June 15 announcement said agent subsidies were ending because Claude Code’s 5-hour reset cost the company an estimated $0.18 per free user per day. OpenAI’s pricing page showed that the free Codex tier allowed 50 agent turns daily with GPT-5.1 Mini only. GitHub’s pricing calculator revealed that agent mode consumed 2.5 credits per request, making a simple multi-file edit cost $0.10 instead of $0.04. Stanford HAI noted that AI coding assistant adoption among developers reached 76 percent in 2026, up from 40 percent in 2024. Free-tier losses became too large to ignore. Major AI coding tools overhauled pricing within the same two weeks.
For developers, the impact was immediate. A solo developer using Copilot Pro for 200 completions and 30 agent edits per day would see monthly costs jump from $10 to about $78, according to the GitHub pricing calculator. Codex users who relied on unlimited code review hit a paywall after 10 reviews per day. The free tier changes meant hobbyists could still test tools, but production workflows required a paid plan. 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.
How Do the Top Options Compare?
| Tool | Pricing Change | Free Tier Impact | Developer Cost Shift | Effective Date |
|---|---|---|---|---|
| GitHub Copilot | Flat $10 Pro replaced by $0.04 per flow credit, agent mode 2.5 credits | 50 completions per day trial only | Solo devs saw $10 to $78 monthly jump for agent workflows | June 12, 2026 |
| OpenAI Codex | Free tier added with 50 agent turns daily, Pro at $20 | 50 turns per day, 10 code reviews, GPT-5.1 Mini only | Unlimited code review ended, Pro required for Max model | June 2, 2026 |
| Cursor | Free chat limited to 20 messages per day, Claude Sonnet 4.5 removed | 20 messages per day, no agent runs | Hobbyists lost agent mode, Pro at $20 | June 9, 2026 |
| Claude Code | Flat rate replaced by shared credit pool, free credits halved | 5-hour reset with 50 credits instead of 100 | Agent runs 30 credits, long sessions costly | June 15, 2026 |
| Gemini Code Assist and Windsurf | Gemini free shut down, Windsurf agent runs halved | No free Gemini tier, Windsurf 5-hour reset | Google users faced $22.80 Standard plan | June 5 and June 9, 2026 |
Prices and limits verified from vendor pricing pages and changelogs as of June 16, 2026. Free tier limits may vary by region and account age.
1. GitHub Copilot , Team agentic coding with large codebases
GitHub Copilot moved to usage-based billing on June 12, 2026, according to the official GitHub changelog. The previous $10 per month Pro plan with unlimited completions disappeared. The new pricing charged $0.04 per flow credit, and agent mode consumed 2.5 credits per request. A multi-file edit that previously cost $0.04 now cost $0.10. For developers who ran daily agent workflows, the monthly bill climbed sharply.
The backlash was immediate. Developers reported that a single refactor across five files burned 18 credits, or $0.72. GitHub’s pricing calculator showed a solo developer making 200 completions and 30 agent edits per day would pay about $78 per month, up from $10. The community thread on the GitHub Copilot hidden costs backlash drew thousands of comments within 48 hours. Many users pointed to the hidden cost of agent mode being 2.5x more expensive than standard chat.
For teams, the shift was not all negative. The new model allowed per-seat billing that matched actual usage, and GitHub introduced a $39 per month Team plan with a 20 percent discount on flow credits. But the free tier was gutted. New users received only 50 completions per day during a 14-day trial. After that, no free Copilot Pro access remained. The change forced many open-source maintainers to switch to local models or free alternatives.
Key strengths:
- ✅ Per-usage pricing gave clearer cost visibility for large teams
- ✅ Team plan offered 20 percent discount on flow credits
- ✅ Agent mode quality remained high with GPT-5.1 integration
- ❌ Agent mode multiplier of 2.5x made refactors expensive
- ❌ Free trial limited to 50 completions per day
- ❌ No free Pro tier after trial
Who it’s for: Teams that can track per-seat usage and need agentic coding on large repositories.
2. OpenAI Codex , Free-tier agentic coding with strict daily caps
OpenAI added a free tier to Codex on June 2, 2026, but the limits were tight. The free plan allowed 50 agent turns per day with GPT-5.1 Mini only. Code review was capped at 10 reviews per day. Pro users at $20 per month received 200 agent turns and access to GPT-5.1 Max. The free tier was open to all ChatGPT accounts, including those in the ChatGPT free tier with ads.
Developers who relied on Codex for unlimited code review hit a paywall after 10 reviews on the free plan. The pricing page showed that the Pro plan lifted the review cap to 100 per day. For hobbyists, the free tier was enough for small bug fixes, but not for full-featured agent workflows. OpenAI positioned the free Codex as a way to test the tool before paying.
The competitive context was clear. OpenAI wanted to push free users toward ChatGPT Pro while also competing with GitHub Copilot’s usage-based model. The free Codex tier excluded the new GPT-5.1 Max model, which was reserved for Pro and API customers. Independent developers noted that 50 turns per day was roughly one hour of active coding.
Key strengths:
- ✅ Free tier gave 50 agent turns daily without payment
- ✅ Pro plan doubled agent turns to 200 for $20 monthly
- ✅ GPT-5.1 Max available on Pro while Mini stayed free
- ❌ Code review cap of 10 per day on free plan
- ❌ Free tier locked to GPT-5.1 Mini, not Max
- ❌ Unlimited code review ended for previous free users
Who it’s for: Hobbyists and students who want to test agentic coding without paying, but can live with strict daily caps.
3. Cursor , Fast free chat but no agent runs
Cursor tightened its free tier on June 9, 2026. The free plan limited AI chat to 20 messages per day and removed Claude Sonnet 4.5 from the free model list. Agent runs, which allowed multi-file edits, were removed entirely from the free plan. Paid Pro at $20 per month restored 500 fast requests and agent access.
The change angered developers who used Cursor for free daily coding. Many had built workflows around Claude Sonnet 4.5 in the free tier. After June 9, they saw a paywall. The Cursor team said the free tier was never meant for production use, but users pointed out that 20 messages per day made even small tasks impossible. The 20-message cap also applied to code generation requests, not just chat.
For students and open-source maintainers, Cursor’s free tier became a demo. The Pro plan remained popular among professional developers because of fast indexing and deep codebase understanding. But the gap between free and paid widened sharply in June 2026.
Key strengths:
- ✅ Pro plan kept 500 fast requests and full agent mode for $20
- ✅ Fast indexing and codebase search remained best in class
- ✅ Clear cutoff date of June 9 gave users notice
- ❌ Free chat capped at 20 messages per day
- ❌ Claude Sonnet 4.5 removed from free tier
- ❌ No agent runs on free plan
Who it’s for: Professional developers who will pay $20 monthly for agentic coding and fast index.
4. Claude Code , Agentic coding with credit pool replacing flat rate
Anthropic ended the Claude Code flat-rate subsidy on June 15, 2026. The previous $5 per month plan with 5-hour reset limits vanished. In its place, Anthropic introduced a shared credit pool across Claude Pro and Claude Code. The free tier kept a 5-hour reset but with 50 percent fewer credits. The official announcement said the change was necessary because agentic coding cost 10x more than standard chat.
The credit pool meant developers paid from one balance for both Claude chat and Claude Code agent runs. A typical agent run consumed 30 credits, while a standard chat consumed 2. The Pro plan at $20 monthly included 1,500 credits, forcing users to budget carefully. Free users saw their 5-hour reset drop from 100 credits to 50, enough for one short agent run.
Developers reacted with frustration. Many had relied on the flat rate for extended debugging sessions. The new model made long agentic coding sessions expensive. Anthropic said the credit pool would reset monthly for paid users, but free users still had the 5-hour reset. The shift pushed some developers to try local models like Mistral Vibe or Grok Build.
Key strengths:
- ✅ Shared credit pool gave predictable monthly budget for paid users
- ✅ Free tier kept 5-hour reset, just smaller
- ✅ Claude Opus 4.8 remained available on Pro
- ❌ Agent runs consumed 30 credits, making long sessions costly
- ❌ Free tier credits halved from 100 to 50 per reset
- ❌ Flat-rate subsidy ended with no replacement
Who it’s for: Developers who use Claude for both chat and coding and want a single credit budget.
5. Gemini Code Assist and Windsurf , Comparing free tier shutdowns and halved limits
Google shut down Gemini Code Assist’s free tier entirely on June 5, 2026. The product moved to a single Standard plan at $22.80 per month, with no free option. The change affected individual developers and students who had used the free tier for inline code suggestions. Google’s AI blog said the shutdown was necessary to focus on enterprise support for large codebases.
Meanwhile, Windsurf changed its free tier on June 9. Agent runs were cut in half, and the reset window moved from 3 hours to 5 hours. Free users could still run basic completions, but agentic features required the Pro plan at $15 monthly. Zed also removed its free AI inline completion beta on June 10, pushing users to Zed AI Pro.
The combined effect was that developers who used free tiers across multiple tools had no reliable free option left. A developer who used Gemini Code Assist for free, Windsurf for free agent runs, and Zed for free completions lost all three within one week. The only remaining free agentic coding options were OpenAI Codex’s 50-turn cap and Claude Code’s halved credits. For many, the free ride ended in June 2026.
Key strengths:
- ✅ Google focused on enterprise support with a clear paid plan
- ✅ Windsurf Pro at $15 remained cheaper than Copilot
- ✅ Zed Pro offered one-time license option
- ❌ Gemini Code Assist had no free tier after June 5
- ❌ Windsurf agent runs halved on free plan
- ❌ Multiple free tier losses in one week
Who it’s for: Developers evaluating whether to pay for a single specialized tool or abandon free tiers altogether.
Frequently Asked Questions
What changed with GitHub Copilot pricing in June 2026?
GitHub Copilot replaced its flat $10 monthly Pro plan with a usage-based model on June 12, 2026. The new pricing charged $0.04 per flow credit, and agent mode used 2.5 credits per request. Agentic workflows became significantly more expensive for daily users.
Did OpenAI Codex offer a free tier in June 2026?
Yes. OpenAI added a free Codex tier on June 2, 2026, but it was limited to 50 agent turns per day and 10 code reviews. The free tier used GPT-5.1 Mini only. Pro users at $20 monthly received 200 turns and GPT-5.1 Max.
Why did Anthropic end the Claude Code flat rate?
Anthropic ended the flat-rate subsidy on June 15, 2026, because agentic coding cost the company significantly more than standard chat. The shared credit pool was designed to align pricing with actual usage. Free tier credits were halved from 100 to 50 per 5-hour reset.
Which free AI coding tool shut down in June 2026?
Google shut down Gemini Code Assist’s free tier on June 5, 2026. The product moved to a single Standard plan at $22.80 per month. Zed also ended its free AI inline completion beta on June 10, 2026.
How much did a solo developer's Copilot bill increase?
A solo developer using 200 completions and 30 agent edits per day saw monthly costs jump from $10 to about $78 under the new usage-based pricing. This estimate came from GitHub’s pricing calculator and was widely reported in developer forums.
Can I still use Cursor for free in June 2026?
Yes, but the free tier was limited to 20 chat messages per day and had no agent runs. Claude Sonnet 4.5 was removed from the free plan on June 9, 2026. Paid Pro at $20 monthly restored 500 fast requests and agent access.
What was the cheapest paid AI coding tool in June 2026?
Windsurf Pro at $15 monthly was the cheapest paid option with agentic features. Cursor Pro and OpenAI Codex Pro both cost $20 monthly. GitHub Copilot’s Team plan started at $39 monthly with a 20 percent discount on flow credits.
What Should You Remember?
- Usage-based billing replaced flat rates across GitHub Copilot and Claude Code in June 2026.
- GitHub Copilot Pro moved from $10 to per-credit pricing on June 12, with agent mode costing 2.5 credits per request.
- OpenAI Codex added a free tier with 50 agent turns daily but locked the free plan to GPT-5.1 Mini.
- Cursor capped free chat at 20 messages and removed agent runs entirely from the free plan on June 9.
- Anthropic halved Claude Code free credits and replaced the flat rate with a shared credit pool on June 15.
- Google shut down Gemini Code Assist free tier on June 5, leaving no free option for inline suggestions.
- Solo developers saw monthly costs jump from $10 to about $78 for the same Copilot agent workflows.
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.
