OpenAI on June 26 previewed GPT-5.6, a three-tier model family designated Sol, Terra, and Luna, in a gated rollout restricted to roughly 20 partner organizations after the company shared its release plans with the U.S. government (OpenAI, VentureBeat). Sol targets the hardest problems including complex coding and cybersecurity research, Terra handles high-volume business workflows, and Luna serves as a fast, low-cost tier for everyday automation (OpenAI). Token pricing runs at $5 input and $30 output per million for Sol - matching the current GPT-5.5 flagship - $2.50/$15 for Terra, and $1/$6 for Luna. The family also revises prompt caching with explicit cache breakpoints, a 30-minute minimum cache lifetime, a 1.25x surcharge on cache writes, and a 90% discount on cache reads, making cost modeling more predictable for long-running agent pipelines (OpenAI Help Center). Broader API availability is planned for the weeks ahead.
June 30 closes the first full billing cycle under GitHub Copilot’s usage-based model, which replaced flat-rate subscriptions on June 1, 2026 (GitHub Blog). Plans now bundle monthly AI Credits priced at $0.01 per credit: Pro ($10/month) includes $15 in credits, Pro+ ($39/month) $70, and Max ($100/month) $200, with Business and Enterprise customers receiving promotional credit boosts through September 1 (GitHub Docs). Developer reception since launch has been sharply negative: one Pro+ subscriber reported consuming roughly 8% of the 7,000-credit monthly allotment within two hours of agentic coding sessions, attributing the drain to multi-step agent workflows that generate far more tokens than the single-turn prompts the prior flat-rate model was sized around (TechCrunch, gHacks).
Google delayed the general availability of Gemini 3.5 Pro past its announced June target, pushing the launch to July after reviewing feedback from limited Vertex AI enterprise testing (Bind AI). Sundar Pichai had committed to a June GA window at Google I/O on May 20, with planned specifications that include a 2-million-token context window and a Deep Think reasoning mode (Google Blog). The delay coincides with a significant loss of senior research talent: Noam Shazeer, Gemini co-lead and co-author of the original Transformer paper, departed for OpenAI on June 18 (CNBC), while AlphaFold project lead John Jumper and researchers Jonas Adler and Alexander Pritzel left for Anthropic within the same period (TechCrunch, Fortune).