OpenAI has begun a limited preview of the GPT‑5.6 family: **Sol** (flagship), **Terra** (balanced, 2× cheaper than GPT‑5.5), and **Luna** (fast, low‑cost). Sol introduces a new “max” reasoning mode and an “ultra” mode that uses sub‑agents. Early benchmarks show state‑of‑the‑art performance on coding (Terminal‑Bench 2.1), biology (GeneBench v1), and cybersecurity (ExploitBench, ExploitGym). The models ship with OpenAI’s most robust safety stack to date, but the preview may block or delay some requests.
GLM‑5.2 is Z.ai’s newest open‑source model designed for long‑horizon coding tasks. It supports a solid 1 million‑token context, introduces flexible “effort” levels for balancing speed and capability, and uses the IndexShare architecture to cut per‑token FLOPs by 2.9×. Benchmarks show it outperforms its predecessor (GLM‑5.1) and ranks as the strongest open‑source coding model, closing the gap to leading closed‑source systems.
Z.ai: announced GLM‑5.2, a frontier‑intelligence model with open weights, a 1 M‑token context window and two reasoning effort levels for coding and agentic tasks. OpenCode: reported that GLM‑5.2 has risen to 6th place on their leaderboard within three days of release. ollama: highlighted GLM‑5.2 as the strongest open‑source coding model yet, now available on Ollama’s US cloud powered by NVIDIA AI Blackwell GPUs. GitHub Changelog: warned that Opus 4.6 (fast) will be deprecated in all Copilot experiences on June 29 2026 and urged migration to Opus 4.8 (fast). Cursor: introduced a new /automate skill that lets agents set up automations from plain‑language task descriptions, configuring triggers, instructions and tools automatically.
Kimi.ai: Released the open‑source Kimi‑K2.7‑Code model, reporting +21.8% on Kimi Code Bench v2, +11.0% on Program Bench, +31.5% on MLS Bench Lite, and 30% lower reasoning overthinking. Google Research: Launched Gemini‑SQL2, a text‑to‑SQL system built on Gemini 3.1 Pro that hits state‑of‑the‑art scores on the BIRD benchmark. OpenAI: Added saved Codex rate‑limit resets for Go, Plus, Pro, and Business tiers (one free reset) and a two‑week invite program letting Plus/Pro users earn extra resets by inviting friends. Claude: Made dynamic workflows in Claude Code generally available, letting the model orchestrate parallel sub‑agents for complex tasks like codebase‑wide bug hunts and verify work before returning results.
Claude Fable 5 is Anthropic’s latest Mythos‑class model, released with safety classifiers that route risky queries to Claude Opus 4.8. It claims state‑of‑the‑art performance on coding benchmarks, dramatically faster codebase migrations, and higher token efficiency than previous Claude models.
NVIDIA’s Nemotron‑3 Ultra is a frontier‑scale LLM with 550 B total (55 B active) parameters, a hybrid LatentMixture‑of‑Experts (LatentMoE) architecture, and up to 1 M token context length. It runs on NVIDIA GPUs (minimum 4 × B200/H100) and offers configurable reasoning traces, multi‑token speculative decoding, and multilingual support. Benchmarks show strong performance on agentic, reasoning, and long‑context tasks.
NVIDIA AI announced the release of Nemotron 3 Ultra, a 550 B MoE model that speeds inference fivefold, lowers agentic task costs up to 30 % and excels at coding, deep research, and long‑horizon planning. OpenCode noted that Nemotron 3 Ultra is now free with 1 M context and fully open source. Ollama said the model is available on its cloud platform, offering launch commands for Claude, Hermes and OpenClaw. OpenAI introduced a new memory system for ChatGPT that automatically tracks important details, doubles memory capacity for Plus and Pro users in the US, and lets users review and steer remembered content via a summary. Cursor added an interactive context‑usage report in its canvas, breaking down token distribution across prompts, tools, rules and skills.
Google Gemma: Announces Gemma 4 12B, a unified encoder‑free multimodal model for laptops released under an Apache 2.0 license. Google AI Developers: Highlights that Gemma 4 12B bridges their mobile E4B and larger 26B MoE models, offering frontier‑class reasoning and native audio. NVIDIA: Notes local AI agents advancing on DGX Spark and RTX PCs, with OpenShell arriving on Windows, new agentic AI optimizations, Broadcast 2.2, and upcoming RTX acceleration for Adobe apps and Blender. Visual Studio Code: Reports May updates—Agents window now stable, BYOK with air‑gapped support, and an integrated browser that can emulate devices and preview HTML without extensions. Ideogram: Introduces Ideogram 4.0, an open image model with downloadable weights, fine‑tuning on personal data, and availability across all plans and the API.
Gemma 4 12B is Google’s new 12‑billion‑parameter multimodal model that runs locally on consumer laptops (≈16 GB VRAM). It eliminates separate vision and audio encoders, delivers reasoning close to the larger 26 B Mixture‑of‑Experts model, and is released under an Apache 2.0 license with full tool‑chain support.
Qwen 3.7‑Plus is a new multimodal agent model that adds vision capabilities to the strong text backbone of Qwen 3.7. It can read screens, interact with GUIs, and generate code from visual references while keeping the coding and tool‑use strengths of its predecessor. Benchmarks show notable gains in several coding‑related tasks, especially in terminal‑based and spreadsheet benchmarks.
MiniMax released its latest M‑series model, **MiniMax‑M3**, on June 1 2026. The model is marketed for agentic reasoning, tool use, coding, multimodal chat input, and long‑context tasks. It follows a series of MiniMax models (M2.5, M2.1) that already claimed state‑of‑the‑art (SOTA) performance in programming, code refactoring, and tool calling.
