All GitHub Copilot plans moved to usage-based billing on June 1, replacing unlimited request access with a GitHub AI Credits system where consumption is metered against input, output, and cached tokens at per-model API rates (GitHub Blog). Copilot Pro+ includes $39 in monthly AI Credits at $39 per month, Copilot Business provides $19 per seat per month with matching credits, and Copilot Enterprise $39 per seat; additional credits can be purchased once the allotment is exhausted (GitHub Docs). Developer response was sharply negative - community posts described unexpectedly rapid credit depletion, and screenshots of projected overages circulated on forums within hours of the transition going live (Windows Forum). Visual Studio Magazine reported that developers characterized the change as ending predictable flat-rate pricing in favor of a consumption model that transfers cost uncertainty to the user, with one commenter summarizing the shift as “you will get less, but pay the same price” (Visual Studio Magazine).
MiniMax released M3 on June 1, an open-weight multimodal model targeting long-horizon coding-agent workflows with a 1 million token context window and native support for text, image, and video inputs (MiniMax); its MSA (MiniMax Sparse Attention) architecture delivers more than 9x prefill speedup and 15x decoding speedup at 1M-token context versus the prior M2, at approximately 1/20th the per-token compute (MarkTechPost). The company-reported SWE-Bench Pro score of 59.0% is cited as exceeding results from GPT-5.5 and Gemini 3.1 Pro at API pricing 5-10% of those models’ rates, though open weights and an independent technical report were committed for release within 10 days of launch rather than at the time of announcement (VentureBeat, MiniMax). Separately, Google is removing the legacy Gemini Interactions API response schema on June 8 - teams still relying on the outputs array must migrate to the typed steps array or add the Api-Revision: 2026-05-07 header, which stops functioning on that date (Google AI for Developers, Byteiota). The MCP maintainers also published the 2026-07-28 specification release candidate on May 21, introducing a stateless protocol layer that removes the Mcp-Session-Id header and allows MCP servers to run behind standard round-robin load balancers without shared session stores; the final specification is scheduled for July 28 (MCP Blog).