OpenAI Unveils GPT-5.6 Series With Sol, Terra, and Luna in Limited Preview
OpenAI launched a limited preview of its GPT-5.6 model family on June 26, 2026, featuring Sol, Terra, and Luna. The flagship Sol model set a new record on Terminal-Bench 2.1 for agentic coding. Access is currently restricted to roughly 20 government-vetted partner companies.
OpenAI and Broadcom Unveil Jalapeño, a Custom Chip Built From Scratch for LLM Inference
OpenAI and Broadcom on June 24, 2026 unveiled Jalapeño, OpenAI's first custom AI accelerator built specifically for large language model inference. Developed in just nine months, the chip promises roughly 50% lower inference costs compared to current GPUs and is slated for initial deployment by end of 2026.
xAI's Grok Imagine Video 1.5 Tops AI Video Leaderboard at a Fraction of Sora 2's Price
xAI launched Grok Imagine Video 1.5 on June 17, 2026, instantly claiming the #1 position on the Artificial Analysis Image-to-Video Arena leaderboard. The model generates synchronized audio and video in a single pass, priced at $4.20 per minute — 86 percent below OpenAI's now-deprecated Sora 2 Pro.
.png&w=3840&q=75)
Trump Administration Forces Anthropic to Pull Two Flagship AI Models, Citing National Security
The US Commerce Department gave Anthropic a 90-minute ultimatum to restrict its Claude Fable 5 and Mythos 5 models over alleged access granted to a China-linked firm. Anthropic chose to pull the models entirely; a June 15 White House meeting ended without resolution.
Zhipu AI Releases GLM-5.2 With 1M-Token Context Window Under MIT License
Chinese AI lab Zhipu AI announced GLM-5.2 on June 13, 2026 — a 744-billion-parameter open-source model featuring a usable 1M-token context window released under the MIT license. Designed for long-horizon agent tasks and coding workflows, the model's public API is set to launch within the week.
Moonshot AI Releases Kimi K2.7-Code, Claims to Surpass Claude Opus 4.8 on Tool-Use Benchmark
Moonshot AI released Kimi K2.7-Code on June 12, 2026, an open-weight coding model with 1 trillion total parameters. The model posts a 21.8% gain over K2.6 on Kimi Code Bench v2 and cuts reasoning-token usage by roughly 30%, according to company benchmarks.
Google releases DiffusionGemma, an open model that generates text up to 4x faster
Google released DiffusionGemma on June 10, 2026, an experimental open model under Apache 2.0. The 26B model generates text in parallel blocks rather than word by word, delivering up to 4x faster inference on dedicated GPUs.
Anthropic Opens Its Most Powerful Model to the Public With Claude Fable 5
Anthropic has launched Claude Fable 5, the first publicly available model in its Mythos tier, which sits above the Opus class. The model ships with strict safeguards that block responses in high-risk areas such as cybersecurity and biology.