On June 12, 2026, Google cut Gemini 2.5 Pro API input pricing by 40 percent. Output prices fell 25 percent. The Google AI blog confirmed the new rates on its official site. This was the second major Google cut in three months. It put direct pressure on OpenAI and Anthropic. The move made Gemini Pro cheaper than GPT-4.1 for many workloads. Developers saw input costs drop to $0.70 per million tokens. Output costs fell to $2.80 per million. Google also tightened its free tier request limit to five requests per minute. The company said the cuts were part of a broader efficiency push. But the tighter free tier frustrated smaller projects. Many developers called the timing deliberate. It followed a brutal May 2026 price war. Google wanted the top spot in unit cost. That strategy worked for a week. This was part of a broader Google price cut strategy.
On June 18, 2026, OpenAI updated its official pricing page. The company cut GPT-4.1 Mini input prices by 25 percent. Input tokens fell from $0.40 to $0.30 per million. Output prices dropped to $1.20 per million. OpenAI also launched a new Batch API. Batch jobs got a 50 percent discount for 24 hour asynchronous processing. That undercut Google’s batch offer by five percentage points. Free tier users faced a new rate limit of five requests per minute. Some developers complained the free tier now failed for small apps. OpenAI said the changes rewarded high volume users. The OpenAI pricing page was the named source. This followed weeks of speculation about an OpenAI price response. It came 10 days after Anthropic’s credit pool change. The combined moves reset developer expectations. For budget conscious teams, GPT-4.1 Mini became the cheapest model for batch inference. The changes were covered in the AI API free tier limits report and the major AI API pricing model updates story.
On June 15, 2026, Anthropic ended its flat-rate agent API subsidy. The official Anthropic changelog replaced flat access with a monthly credit pool. Free tier users received 500,000 credits. Credits reset every five hours. Paid users moved to per token billing. The change hit developers who ran long agent loops on the free plan. It also hit small startups that had built workflows on unlimited token allowance. Anthropic framed the change as a fairness fix. But many users called it a paywall. The move followed a May 2026 report that agentic billing was out of control. Anthropic said the credit pool would prevent abuse. It also introduced a new paid tier with priority access. The change took effect immediately at 2 p.m. Pacific. The full story appeared in Anthropic ends agent subsidy.
Why it mattered. Independent analysis from Stanford HAI found API list prices fell 31 percent year over year as of June 2026. But free tier rate limits grew stricter across Google, OpenAI, and Mistral. The gap between cheap paid access and constrained free access widened. Google, OpenAI, and Anthropic all used June to push serious developers toward paid plans. Casual users lost headroom. The AI Index report called the pattern a two speed market. Budget teams gained from price cuts. Hobbyists lost from rate limits. That tension defined June 2026 AI API pricing. Our AI free tier shifts report tracked the access changes in real time.
How Do the Top Options Compare?
| Provider | June 2026 Change | New Price or Limit | Affected Users | Effective Date |
|---|---|---|---|---|
| Google Gemini API | 40% input price cut on Gemini 2.5 Pro | $0.70 per million input tokens | All API developers | June 12, 2026 |
| OpenAI API | 25% price cut on GPT-4.1 Mini plus Batch API discount | $0.30 per million input tokens | All API developers | June 18, 2026 |
| Anthropic Claude API | Flat-rate agent access replaced by credit pool | 500,000 credits per month free tier | Agent developers and free-tier users | June 15, 2026 |
| Mistral AI API | Open-weight model release with free tier cap | 10 requests per minute free tier | Free API developers | June 20, 2026 |
Prices are per million tokens unless stated. Free-tier request limits may vary by region and model version.
1. Google Gemini API , Best for Developers Tracking Unit Costs
Google cut Gemini 2.5 Pro API input prices by 40 percent on June 12, 2026. The official Google AI blog confirmed the new rate of $0.70 per million input tokens. Output prices fell to $2.80 per million tokens. That made Gemini Pro cheaper than GPT-4.1 Mini on many long context tasks.
The cut followed Google’s May 2026 decision to put Gemini 3.5 Flash on the free tier. But Google also tightened free tier compute quotas. Free users now get five requests per minute. Paid users saw no rate limit change. The company framed the move as a win for high volume developers. Google’s AI price cuts put OpenAI and Anthropic on defense.
For mixed workloads, the new pricing cut typical bills by about 38 percent. Batch processing still required a separate discount. Developers who needed low latency paid full rate. The free tier was no longer viable for small apps. That was a deliberate trade off.
Key strengths:
- ✅ 40 percent lower input prices
- ✅ Cheaper than GPT-4.1 Mini for high volume
- ✅ Gemini 3.5 Flash free tier remains
- ✅ Clear official pricing page
- ❌ Free tier rate limit dropped to five requests per minute
- ❌ Batch discount not automatic
- ❌ Output prices still higher than Mistral
Who it’s for: Developers who run high volume Gemini Pro calls and want predictable unit costs.
2. OpenAI API , Best for Batch Inference and Low Priority Jobs
OpenAI cut GPT-4.1 Mini input prices by 25 percent on June 18, 2026. The OpenAI pricing page listed the new input rate at $0.30 per million tokens. Output prices fell to $1.20 per million. The company also launched a Batch API with a 50 percent discount for 24 hour asynchronous jobs. That undercut Google’s batch pricing by five percentage points.
The free tier changed at the same time. OpenAI reduced free API requests to five per minute. Developers on GitHub reported failures in small side projects. OpenAI said the free tier was meant for testing, not production. The AI API free tier limits report covered the backlash. Paid developers with committed spend got priority.
For budget conscious teams, the Batch API became the cheapest way to process large datasets. But batch jobs could wait 24 hours. Real time apps paid full price. The changes rewarded patience. They punished interactivity. That was a clear shift in OpenAI’s pricing philosophy.
Key strengths:
- ✅ 25 percent cheaper GPT-4.1 Mini
- ✅ 50 percent batch discount
- ✅ Broad model lineup remains
- ✅ Pay as you go pricing
- ❌ Free tier cut to five requests per minute
- ❌ Batch jobs have 24 hour latency
- ❌ No price cut for flagship GPT-4.1
Who it’s for: Teams that can batch non urgent inference and want the lowest OpenAI unit cost.
3. Anthropic Claude API , Best for Agent Developers Who Accept Metered Access
Anthropic ended its flat rate agent API subsidy on June 15, 2026. The Anthropic changelog replaced flat access with a 500,000 credit pool for free tier users. Credits reset every five hours. Paid users moved to per token billing with priority access. This hit developers who ran long agent loops on the free plan.
The change followed months of reports about agentic API abuse. Anthropic said abuse drove the decision. Free tier users lost unlimited token allowance. Small startups called it a paywall. The Anthropic agent subsidy end story detailed the backlash.
Independent analysis from Stanford HAI noted Anthropic unit costs fell, but access tightened. The credit pool gave clearer accounting. It also made agent calls expensive for power users. Developers who needed long contexts now paid per token. That changed the economics of agentic coding.
Key strengths:
- ✅ Clear credit accounting
- ✅ Paid users get priority access
- ✅ 500,000 free credits still useful
- ✅ Reset every five hours prevents abuse
- ❌ Free flat rate access ended
- ❌ Per token billing raises agent costs
- ❌ Five hour reset confuses users
Who it’s for: Developers building agentic workflows who can budget for credits and want predictable abuse controls.
4. Mistral AI API , Best for Open Weight Model Tinkerers
On June 20, 2026, Mistral released a new open weight model through its API. The official site published weights and pricing. Free tier users got 10 requests per minute with no credit card required. That was more generous than Google’s five requests but still tight.
The model undercut Google and OpenAI on small model API pricing. Full weights went to Hugging Face. Developers could self host and avoid API costs entirely. The open source release matched Mistral’s June pattern of free tier adjustments. Mistral Vibe free tier covered the consumer side.
Mistral’s API lacked enterprise support and long context windows. But for developers who wanted control, it was the cheapest option. No vendor lock. No hidden usage multipliers. The free tier limit meant production use required payment. Still, the open weights meant anyone could run the model locally.
Key strengths:
- ✅ Open weights available
- ✅ Cheapest small model API
- ✅ No credit card for free tier
- ✅ No vendor lock
- ❌ Free tier capped at 10 requests per minute
- ❌ Smaller context window than GPT-4.1
- ❌ Limited enterprise support
Who it’s for: Developers who want open weight models and self hosting options without vendor lock.
Frequently Asked Questions
What was the biggest AI API price cut in June 2026?
Google cut Gemini 2.5 Pro input prices by 40 percent on June 12. The new rate was $0.70 per million tokens. Output prices fell 25 percent to $2.80 per million.
Did OpenAI change API prices in June 2026?
Yes. OpenAI cut GPT-4.1 Mini input prices 25 percent on June 18. It also introduced a Batch API with a 50 percent discount for 24 hour jobs.
What happened to Anthropic's Claude API free tier?
Anthropic ended flat rate agent access on June 15. Free users now receive 500,000 credits that reset every five hours. Paid users moved to per token billing.
Which providers tightened free tier API limits in June 2026?
Google, OpenAI, and Mistral all tightened limits. Most free tiers now allow five to ten requests per minute. Anthropic replaced unlimited flat access with credits.
Which AI API had the cheapest open weight model?
Mistral AI released a new open weight model on June 20. Its API undercut Google and OpenAI for small model tasks. Weights are available on Hugging Face.
Where can I verify these June 2026 pricing changes?
Check the Google AI blog, OpenAI pricing page, Anthropic changelog, and Mistral AI site. Stanford HAI also published independent price trend analysis.
What Should You Remember?
- Google cut Gemini 2.5 Pro input prices 40 percent on June 12, lowering input to $0.70 per million tokens.
- OpenAI cut GPT-4.1 Mini by 25 percent and added a 50 percent Batch API discount on June 18.
- Anthropic ended flat rate agent access on June 15, replacing it with a 500,000 credit pool.
- Free tier API limits got tighter across Google, OpenAI, and Mistral, falling to five or ten requests per minute.
- Mistral released an open weight model on June 20 with a free tier capped at 10 requests per minute.
- Stanford HAI data showed API list prices fell 31 percent year over year, but access controls tightened.
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.
