AI Daily｜Mistral Large 4 Released, OpenAI Launches Decisions API, and Google Debuts EmbeddingGemma 2
Model Releases & Updates
Mistral Large 4 (‘Le Chonk’) — Mistral AI
- Bottom line：Mistral AI introduced Mistral Large 4, a 1-trillion parameter natively multimodal open-weight model with 49 billion active parameters.
- Architecture & Specs：Features 1 trillion total parameters with 49 billion active parameters, supporting a 1-million token context window and up to 256K output tokens.
- Performance & Workloads：Delivers state-of-the-art results on critical enterprise workloads including cyber defense, manufacturing, and finance.
- Availability：Available via API immediately, with open weights scheduled for release at the end of October.
- Source：Official Announcement
EmbeddingGemma 2 — Google DeepMind
- Bottom line：Google DeepMind released EmbeddingGemma 2, a 740M-parameter open multimodal embedding model mapping text, code, images, audio, and video into a unified vector space.
- Multimodal Capabilities：Natively maps text, code, image, audio, and video modalities into a shared vector space.
- Open Source：Built on the Gemma 4 architecture and released under an Apache 2.0 license with weights available on Hugging Face and Kaggle.
- Efficiency：Offers strong performance competitive with models twice its size while remaining lightweight enough for on-device deployment.
- Source：DeepMind Blog
Product Releases & Updates
Decisions API — OpenAI
- Bottom line：OpenAI launched the Decisions API in public beta, enabling applications to classify, score, and route requests up to 10x faster.
- Core Functionality：Accepts text and image inputs to generate probabilities, predefined choices with confidence scores, and numeric evaluation scores.
- Performance：Operates up to 10x faster than GPT-6 Luna through the Responses API for near real-time decision making.
- Use Cases：Widely utilized for model/tool routing, dataset categorization, and flagging risky tool calls.
- Source：
Claude for Google Workspace — Anthropic
- Bottom line：Anthropic launched Claude for Google Workspace in beta, allowing users to interact with and edit Google Docs, Sheets, and Slides directly via a sidebar.
- Integration：Integrates directly into Google Docs, Sheets, and Slides via a side panel to read and edit content in place.
- User Control：Users can review and approve each individual edit before it is applied to the document.
- Availability：Available in beta across all paid Claude plans, respecting existing Google sharing permissions.
- Source：
Figma Agent General Availability — Figma
- Bottom line：Figma moved its AI agent out of beta, bringing improved output quality, better instruction-following, and enhanced long-running task handling.
- Performance Gains：Wins in over 60% of human-graded evaluations conducted with professional designers.
- Task Handling：Enhanced capability for managing longer-running design workflows and adhering strictly to prompt instructions.
- Source：
Industry News
DeepSeek Series B Financing — DeepSeek
- Bottom line：DeepSeek reportedly neared completion of a massive financing round of at least 80 billion yuan (approx. $12 billion), with Tencent and CATL participating.
- Funding Scale：Raised at least 80 billion yuan (approx. $12 billion), significantly exceeding its initial target of 50 billion yuan.
- Key Investors：Backed by major strategic investors including tech giant Tencent and battery manufacturer CATL.
- Future Plans：The company is reportedly targeting an initial public offering (IPO) in early 2027.
- Source：
Cyber Verification Program Expansion — Anthropic
- Bottom line：Anthropic expanded its Cyber Verification Program to grant verified security professionals broader access to its frontier models for defensive and offensive work.
- Model Access：Grants verified cybersecurity professionals access to Claude Mythos 5.1, Opus 5.5, and Sonnet 5.5.
- Authorized Scope：Introduces new access tiers specifically designed to support authorized offensive work such as penetration testing and red-teaming.
- Source：Anthropic Blog
Research Papers
Sharing AI Progress in Mathematics — OpenAI
- Bottom line：OpenAI published a collection of new mathematical results produced by an internal frontier model, alongside Lean formalizations on GitHub.
- Scope of Results：Releases 722 papers organized into 372 groups covering number theory, algebraic geometry, combinatorics, and mathematical physics.
- Verification：Approximately 160 papers include Lean formalizations to ensure computer-verified mathematical rigor.
- Advisory Guidance：The release approach was informed by consultations with an independent Advisory Group on Mathematics and Artificial Intelligence.
- Source：OpenAI Blog
Other Highlights
Rebuilding Git Infrastructure for Agent-Scale Development — GitHub
- Bottom line：GitHub announced a complete reconstruction of its Git infrastructure to handle exponential concurrency spikes driven by AI coding agents and developers.
- Traffic Surge：Monthly Git events reached 473.3 billion between September 2025 and August 2026, doubling year-over-year.
- Agent Contribution：In September alone, coding agents and developers generated 7.38 billion commits, marking a fivefold increase compared to the previous year.
- Live Architecture Overhaul：GitHub is rebuilding foundational components without downtime to support future agent-scale workloads.
- Source：GitHub Blog
