OpenAI GPT-5.6 Series Goes Public: Sol, Terra, Luna Now Available Worldwide
OpenAI officially released the GPT-5.6 series on July 9, making Sol, Terra, and Luna available through ChatGPT, Codex, and the OpenAI API globally. The flagship Sol model introduces two new capability tiers — "max" and "ultra" — where "max" allocates additional inference time for exploring alternative solutions and self-correcting approaches, while "ultra" coordinates four parallel agent instances to tackle complex multi-step tasks with higher token consumption for superior results. Terra is positioned as the balanced daily-work model, and Luna as the fastest, most cost-efficient option.
The series represents OpenAI's most robust safety deployment yet, with the accompanying system card detailing extensive evaluations in biological and cybersecurity domains. GPT-5.6 API pricing varies by tier, with per-million-token rates designed to accommodate everything from lightweight consumer applications to enterprise-grade agentic workflows.
— OpenAI · Xinhua
🔗 OpenAI Research Index · GPT-5.6 Preview Blog · Xinhua Coverage
OpenAI GPT-Live Voice Model: Real-Time Conversation Arrives
On July 8, OpenAI introduced GPT-Live, a new generation of voice models built on a native real-time architecture that fundamentally changes how humans interact with AI. The model supports natural interruptible conversation, pause comprehension, real-time translation, and dictation, while seamlessly orchestrating backend models like GPT-5.5 for complex reasoning and web search. Two versions are available: GPT-Live-1 for Go, Plus, and Pro subscribers, and GPT-Live-1 mini for free users.
OpenAI revealed that over 150 million people now use ChatGPT Voice, Dictation, and related speech features weekly. Product lead Atty Eleti positioned the launch as the beginning of a shift where "voice becomes the primary way we interact with computing devices." The model is rolling out across web, iOS, and Android, with API access planned for the coming weeks.
— OpenAI · Wall Street Journal
🔗 OpenAI GPT-Live Announcement · WSJ Coverage
Meta Muse Spark Enters Coding Arena, Llama API Shuts Down
Meta marked a pivotal strategy shift this week. CEO Mark Zuckerberg emerged from a three-year social media hiatus to personally announce Muse Spark, Meta's most powerful agent model, now entering the programming domain. The model is available via public preview on Meta's Model API portal, and early benchmarks show it competitive with frontier coding models. Meta is simultaneously training a larger model codenamed Watermelon, which has reportedly matched GPT-5.5 on key benchmarks.
In a related move, Meta shut down the Llama API service on July 6, ending its short-lived 14-month experiment in selling API access. The company is pivoting to a dual-track strategy: open-source Llama continues for the community while closed-source Muse powers Meta's proprietary ecosystem across WhatsApp, Instagram, Facebook, and smart glasses. Meta also released Muse Image (codename "Mango"), a generative image model deeply integrated into Instagram and WhatsApp that supports account mention prompts for likeness reuse.
— Meta · The Verge
🔗 Meta Muse Spark Announcement · Llama API Shutdown Coverage
Microsoft Replaces OpenAI and Anthropic Models with In-House MAI
Microsoft has begun a quiet but consequential transition, replacing third-party AI models from OpenAI and Anthropic with its self-developed MAI model family within Excel and Outlook. The new MAI-Thinking 1 model has demonstrated performance matching Claude Opus 4.8 on coding benchmarks, according to internal testing cited by Bloomberg. Tens of thousands of AI prompt requests in these two flagship Office applications are now handled entirely by Microsoft's own models each week.
The shift is driven by cost pressures and data residency requirements. Mustafa Suleyman, Microsoft's AI CEO, is leading the effort to reduce dependency on premium third-party APIs as OpenAI's discounted partnership window narrows. While MAI's overall share of Microsoft's AI inference volume remains modest, the Excel and Outlook deployments represent a beachhead that could expand rapidly across Copilot, Azure AI, and the broader Microsoft 365 ecosystem.
— Bloomberg · Microsoft
🔗 Bloomberg via 163.com · Microsoft AI Blog
NVIDIA and Hugging Face Release Open Data for AI Agents
NVIDIA and Hugging Face jointly announced the Open Data for Agents initiative, publishing over 10 trillion pre-training tokens and millions of post-training samples specifically designed for building AI agents. The release includes region-specific synthetic personas and an interactive Nemotron Post-Training v3 Prompt Atlas, enabling organizations to fine-tune agent models without exposing proprietary data.
On the robotics front, NVIDIA integrated Isaac GR00T 1.7 — a vision-language-action foundation model for humanoid robots — and Isaac Teleop, an open framework for capturing human demonstration data, directly into Hugging Face's open-source LeRobot library. This allows developers to post-train and deploy humanoid robot models through standardized LeRobot workflows. A planned Cosmos 3 world foundation model will further support data generation for robotics when real-world data is scarce.
— NVIDIA · Hugging Face
🔗 Hugging Face Blog - Data for Agents · NVIDIA Isaac GR00T
Z.ai Launches ZCode: Free AI Coding Agent That Beats GPT-5.5 on SWE-Bench
Z.ai (formerly Zhipu AI) launched ZCode, a free "Agentic Development Environment" powered by the GLM-5.2 model, a 744-billion-parameter Mixture-of-Experts architecture. On SWE-bench Pro, GLM-5.2 scored 62.1, surpassing OpenAI's GPT-5.5 at 58.6, though trailing Claude Opus 4.8 at 66.0. On Terminal-Bench 2.1, it scored 81.0 against Claude Opus 4.8's 85.0.
ZCode's pricing is aggressive: the base tier is free, with paid plans starting at $16.20/month (undercutting Cursor Pro at $20). API pricing for GLM-5.2 is $1.40 per million input tokens and $4.40 per million output tokens — a fraction of Claude Opus 4.8's $5/$25. The agent-first IDE supports macOS, Windows, and Linux, and includes remote control via WeChat and Feishu messaging bots, reflecting its Chinese enterprise market focus. ZCode arrives just three weeks after the US suspension of Anthropic's Fable 5 model, creating a strategic opening in the global coding agent market.
— Z.ai · TechTimes
🔗 TechTimes Coverage · EastFrontier Analysis
Mistral AI Releases Leanstral 1.5: Open-Source Formal Verification Powerhouse
Mistral AI released Leanstral 1.5, a 119-billion-parameter sparse Mixture-of-Experts model under the Apache-2.0 license, specializing in formal verification and mathematical theorem proving. The model saturates miniF2F (solving effectively all problems), solves 587 out of 672 PutnamBench problems, and achieves new state-of-the-art scores on FATE-H (87%) and FATE-X (34%) algebra verification benchmarks.
Beyond synthetic benchmarks, Leanstral 1.5 demonstrated practical impact by discovering five previously unknown bugs across 57 real-world open-source repositories, including Rust codebases. The model activates only ~6 billion parameters per token, making it deployable at a fraction of the compute cost its total parameter count would suggest. It supports a 256,000-token context window and is available for free via Mistral's Labs tier API, console playground, and Vibe agent environment.
— Mistral AI · The Agent Times
🔗 Mistral AI Blog · The Agent Times · Hugging Face Model
Next digest: July 12, 2026 — KD Agentic
Top comments (0)