AI News June 1 2026 - Copilot Billing Just Changed, GPT-4.5 Has 26 Days Left, Build Starts Tomorrow

1. GitHub Copilot AI Credits Are Live Today - What Every Developer Needs to Know

As of June 1, 2026, every GitHub Copilot plan has switched from the Premium Request Unit (PRU) system to token-based AI Credits. The flat-rate billing model - where a quick chat question and a 20-step agentic session cost the same - is gone. From today, you pay by tokens: input tokens, output tokens, and cached tokens, converted to GitHub AI Credits at 1 credit = $0.01 USD.

The good news: inline code completions and Next Edit Suggestions remain completely free on every plan, including Copilot Free. They never consume credits. The billing risk is in chat, agent sessions, terminal CLI commands, and Actions-based Copilot code reviews - all of which now meter from your monthly credit allowance. Plan prices have not changed: Copilot Pro remains $10/month (1,000 credits included), Pro+ $39/month (3,900 credits), Business $19/seat (2,000 credits per seat), Enterprise $39/seat (4,000 credits per seat).

Business and Enterprise customers get a promotional uplift through August 2026 - $30/seat effective for Business (vs standard $19) and $70/seat effective for Enterprise (vs $39) in credit value. After August, credits return to the standard base amounts. This is the window to understand your team's actual usage before the promotional buffer disappears. Set a hard cap on additional spend in your GitHub billing settings - this makes a surprise bill literally impossible. Check your Billing Overview page at github.com to see your projected monthly credit consumption before the first June bill lands.

The model cost context: a quick GPT-5 mini chat question costs a fraction of a credit. A long Claude Sonnet 4.6 agent session spanning multiple files can cost several credits. Heavy Pro users who ran agentic sessions daily under the old PRU system are the most at risk of running short - the old system charged a flat per-request rate regardless of token volume; the new system charges by actual token consumption, which for long agentic runs is substantially higher. For full pricing guidance, see our dedicated article: Codex Pro Pricing After the Promo Ends ->

2. GPT-4.5 Retires June 27 - Migrate to GPT-5.3 Now

OpenAI confirmed GPT-4.5 will be retired from ChatGPT on June 27, 2026 following a 30-day sunset period. GPT-4o follows on August 26 after a 90-day sunset. Both are replaced by their successor models in the GPT-5.x family. GPT-4.5 conversations will automatically continue on GPT-5.3 after June 27; GPT-4o conversations will continue on the corresponding current model after August 26.

For developers with production API integrations on gpt-4.5 or gpt-4o model strings, these retirements require active migration - the API will return errors after the cutoff date if the specific model string is still hardcoded. The recommended path is to test your integration against GPT-5.3 now, before June 27, to identify any response characteristic changes that need prompt adjustments. OpenAI's model release notes page at help.openai.com is the authoritative reference for current sunset dates across all models.

3. Microsoft Build Starts Tomorrow - Windows Agent Framework, WSL 3, and the Agent Store

Microsoft Build 2026 opens tomorrow, June 2, at Fort Mason Center in San Francisco. Satya Nadella keynotes at 10am PT / 1pm ET. Free online stream at build.microsoft.com. The confirmed announcement list is the most pre-disclosed in Build's history - Windows Agent Framework APIs embedded in the OS shell, WSL 3 with paravirtualized GPU and NPU access for near-native Linux AI performance on Windows, Azure Agent Mesh for federated agent execution across Windows 365 and Azure Arc (GA Q4 2026), and a Windows Agent Store with 85% developer revenue share. Adobe and Zoom are confirmed launch partners.

The Windows Agent Runtime preview - the execution layer agents run on inside Windows - goes to Insider preview during or immediately after Build. Initial support is text-based agents only; multi-modal agent support (screen interaction) follows in subsequent builds. WSL 3 is the announcement most likely to have immediate practical impact: by giving Linux workloads direct GPU and NPU access at near-native speed, it closes the performance gap that has driven AI developers to prefer Mac or Linux. For full coverage of every confirmed Build announcement, see: Microsoft Build 2026: Everything Confirmed ->