{
  "name": "AIDayNews AI Model Release Tracker",
  "publisher": "AIDayNews",
  "url": "https://www.aidaynews.com/tracker",
  "license": "https://creativecommons.org/licenses/by/4.0/",
  "attribution": "Source: AIDayNews AI Model Release Tracker (https://www.aidaynews.com/tracker)",
  "how_to_credit": "Free to reuse, including commercially. Please name AIDayNews and link to https://www.aidaynews.com/tracker wherever you show this data.",
  "updated": "2026-11-21",
  "count": 83,
  "releases": [
    {
      "date": "2026-11-20",
      "company": "Google DeepMind",
      "model": "Gemini 4 Neon",
      "kind": "Text",
      "summary": "Frontier-level intelligence model for software engineering, knowledge work, and science workflows with fast task completion.",
      "availability": "Gemini API, Google AI Studio, Gemini for business, Google AI Pro and Ultra subscribers",
      "source_name": "Google DeepMind",
      "source_url": "https://deepmind.google/blog/introducing-gemini-4-neon/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-gemini4neon"
    },
    {
      "date": "2026-11-19",
      "company": "Reson8",
      "model": "Rumble-2-Flash",
      "kind": "Audio",
      "summary": "0.5B-parameter speech recognition model for 40+ languages with automatic language detection.",
      "availability": "Open source under Apache 2.0 license, available on Hugging Face Hub and via API.",
      "source_name": "HuggingFace Blog",
      "source_url": "https://huggingface.co/blog/rumble-2-flash",
      "tracker_url": "https://www.aidaynews.com/tracker#m-rumble2flash"
    },
    {
      "date": "2026-11-16",
      "company": "Google DeepMind",
      "model": "Gemini 4 Hydrogen",
      "kind": "Text",
      "summary": "Fast execution model for everyday coding and agentic workflows, available across Google AI products and services.",
      "availability": "Google AI Studio, Gemini API, Gemini for business, Gemini app, Google Search, Google AI Pro and Ultra subscribers",
      "source_name": "Google DeepMind",
      "source_url": "https://deepmind.google/blog/introducing-gemini-4-hydrogen/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-gemini4hydrogen"
    },
    {
      "date": "2026-11-15",
      "company": "Mistral",
      "model": "Timestral",
      "kind": "Other",
      "summary": "Family of time-series foundation models in four sizes from 5M to 209M parameters for probabilistic multivariate forecasting.",
      "availability": "Available as a family of four models; fine-tuning available via Mistral Forge.",
      "source_name": "Mistral AI",
      "source_url": "https://mistral.ai/news/timestral/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-timestral"
    },
    {
      "date": "2026-11-12",
      "company": "Google DeepMind",
      "model": "AlphaGenome 2",
      "kind": "Science",
      "summary": "Next-generation DNA sequence model predicting how genetic variants affect specific human cell types.",
      "availability": "Non-commercial research via API; private preview for commercial research",
      "source_name": "Google DeepMind",
      "source_url": "https://deepmind.google/blog/alphagenome-2-deciphering-how-cells-read-dna-to-better-understand-disease/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-alphagenome2"
    },
    {
      "date": "2026-11-11",
      "company": "Tether AI",
      "model": "BotPsy",
      "kind": "Text",
      "summary": "A 0.8B-parameter model for on-device function calling and agentic workflows built on Qwen3.5-0.8B.",
      "availability": "Model safetensors and GGUFs available for download",
      "source_name": "HuggingFace Blog",
      "source_url": "https://huggingface.co/blog/qvac/botpsy",
      "tracker_url": "https://www.aidaynews.com/tracker#m-botpsy"
    },
    {
      "date": "2026-11-10",
      "company": "Google DeepMind",
      "model": "AlphaProof Code",
      "kind": "Code",
      "summary": "AI framework for formal verification that mathematically proves software cryptographic guarantees and security properties.",
      "availability": "Researchers released the specification and proof for the Signal protocol analysis",
      "source_name": "Wired AI",
      "source_url": "https://www.wired.com/story/a-new-ai-tool-from-google-deepmind-formally-verified-the-signal-protocol/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-alphaproofcode"
    },
    {
      "date": "2026-11-09",
      "company": "Anthropic",
      "model": "Claude Fable 5.2 and Claude Mythos 5.2",
      "kind": "Code",
      "summary": "Upgraded models with improvements in coding, reasoning, and vision capabilities, generating outputs up to 30% faster than previous versions.",
      "availability": "Fable 5.2 available to all users. Mythos 5.2 requires application through Life Sciences and Cyber Verification Programs.",
      "source_name": "Anthropic",
      "source_url": "https://www.anthropic.com/claude-fable-and-mythos-5-2",
      "tracker_url": "https://www.aidaynews.com/tracker#m-claudefable52andclaudemythos52"
    },
    {
      "date": "2026-11-08",
      "company": "Dharma AI",
      "model": "Dharma Field Extractor",
      "kind": "Other",
      "summary": "A 2-billion-parameter model that extracts structured JSON fields from any document type.",
      "availability": "",
      "source_name": "HuggingFace Blog",
      "source_url": "https://huggingface.co/blog/Dharma-AI/dharma-field-extractor",
      "tracker_url": "https://www.aidaynews.com/tracker#m-fieldextractor"
    },
    {
      "date": "2026-11-08",
      "company": "NVIDIA",
      "model": "Cosmos 3 Nano",
      "kind": "Video",
      "summary": "Four open-weight checkpoints for interactive world simulation in driving and robotics tasks.",
      "availability": "Open weights under OpenMDW1.1 for commercial and non-commercial use.",
      "source_name": "HuggingFace Blog",
      "source_url": "https://huggingface.co/blog/nvidia/cosmos-interactive-world-simulation",
      "tracker_url": "https://www.aidaynews.com/tracker#m-cosmos3nano"
    },
    {
      "date": "2026-11-07",
      "company": "Hugging Face",
      "model": "Phonon",
      "kind": "Audio",
      "summary": "On-device text-to-speech model running locally without server or network connection.",
      "availability": "Phonon 31m Apache 2.0 free for all uses. Phonon 40m CC BY-NC free for non-commercial.",
      "source_name": "HuggingFace Blog",
      "source_url": "https://huggingface.co/blog/phonon-tts",
      "tracker_url": "https://www.aidaynews.com/tracker#m-phonon"
    },
    {
      "date": "2026-11-07",
      "company": "Google DeepMind",
      "model": "Gemini Omni 1.2 Flash",
      "kind": "Video",
      "summary": "Generates native 15-second videos with improved consistency, multi-turn editing, and 30% lower costs than Omni 1.1.",
      "availability": "Generally available on Gemini API, Google Cloud, Google Vids, and Gemini for business",
      "source_name": "Google DeepMind",
      "source_url": "https://deepmind.google/blog/gemini-omni-12-flash-fast-and-cost-effective-with-longer-clips/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-geminiomni12flash"
    },
    {
      "date": "2026-11-06",
      "company": "Meta",
      "model": "Muse Realtime Avatar and Muse Realtime Voice",
      "kind": "Other",
      "summary": "New AI models powering Muse's digital avatar for video chat and voice call capabilities with animated character expressions.",
      "availability": "Available to all U.S. users on mobile, web, and Mac",
      "source_name": "TechCrunch AI",
      "source_url": "https://techcrunch.com/2026/11/05/meta-says-us-users-can-now-video-chat-with-their-muse-ai-avatar/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-muserealtimeavatarandmuserealtimevoice"
    },
    {
      "date": "2026-11-01",
      "company": "Mistral",
      "model": "Mistral Large 4",
      "kind": "Text",
      "summary": "Multimodal, multilingual Mixture-of-Experts model with 1T total parameters and 1M token context window.",
      "availability": "Released under Mistral Ecosystem Open-Weight License on Hugging Face.",
      "source_name": "HuggingFace Blog",
      "source_url": "https://huggingface.co/blog/mistral-large4",
      "tracker_url": "https://www.aidaynews.com/tracker#m-large4"
    },
    {
      "date": "2026-11-01",
      "company": "Google DeepMind",
      "model": "Gemini 4 Carbon",
      "kind": "Text",
      "summary": "Model for coding and agentic tasks, with a focus on cybersecurity and long-running work.",
      "availability": "Starting this week, subset of cyber defenders; gradual access through Fairwind Program.",
      "source_name": "Google DeepMind",
      "source_url": "https://deepmind.google/blog/gemini-4-carbon-our-next-frontier-model/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-gemini4carbon"
    },
    {
      "date": "2026-10-30",
      "company": "Liquid AI",
      "model": "LFM3",
      "kind": "Other",
      "summary": "New generation of Liquid Foundation Models with text, vision, audio, omni, and decision models in 2B parameter range.",
      "availability": "Available via Hugging Face transformers library for on-device deployment",
      "source_name": "HuggingFace Blog",
      "source_url": "https://huggingface.co/blog/LiquidAI/lfm3",
      "tracker_url": "https://www.aidaynews.com/tracker#m-lfm3"
    },
    {
      "date": "2026-10-28",
      "company": "Google DeepMind",
      "model": "Genie 4",
      "kind": "Video",
      "summary": "World model generating photorealistic interactive worlds with real-time audio, free-form editing, and extended scene consistency.",
      "availability": "Research preview for trusted testers; rolling out to Google AI Ultra subscribers and early access partners",
      "source_name": "TechCrunch AI",
      "source_url": "https://techcrunch.com/2026/10/27/googles-genie-4-world-model-adds-audio-more-realistic-worlds-and-real-time-steering/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-genie4"
    },
    {
      "date": "2026-10-25",
      "company": "Block",
      "model": "Dessa",
      "kind": "Audio",
      "summary": "Full-duplex speech-to-speech model trained entirely on speech that can run locally on consumer GPUs.",
      "availability": "No public access currently; Block seeking safety researchers to test the model.",
      "source_name": "TechCrunch AI",
      "source_url": "https://techcrunch.com/2026/10/24/blocks-new-voice-model-sounds-eerily-human/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-dessa"
    },
    {
      "date": "2026-10-24",
      "company": "IBM",
      "model": "Granite Speech 5.0 2B",
      "kind": "Audio",
      "summary": "New multilingual speech model for transcription and translation with improved accuracy and robustness.",
      "availability": "Apache 2.0 license with support in Transformers, vLLM, llama.cpp, and mlx-audio",
      "source_name": "HuggingFace Blog",
      "source_url": "https://huggingface.co/blog/ibm-granite/granite-speech-5",
      "tracker_url": "https://www.aidaynews.com/tracker#m-granitespeech502b"
    },
    {
      "date": "2026-10-23",
      "company": "Canva",
      "model": "Canva Craft",
      "kind": "Image",
      "summary": "Image generation model that interprets prompts and generates images up to 2K resolution with accurate text and details.",
      "availability": "Available to all Canva platform users via Dream Lab tab and Canva API",
      "source_name": "TechCrunch AI",
      "source_url": "https://techcrunch.com/2026/10/22/canva-releases-its-first-image-generation-model/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-craft"
    },
    {
      "date": "2026-10-22",
      "company": "Recursive",
      "model": "Flint",
      "kind": "Code",
      "summary": "A coding model that runs on laptops and was built largely with AI assistance from the Eureka Machine.",
      "availability": "Free download on Hugging Face under Apache 2.0 open source license",
      "source_name": "TechCrunch AI",
      "source_url": "https://techcrunch.com/2026/10/21/recursive-launches-flint-a-laptop-coding-model-it-says-was-largely-built-by-ai/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-flint"
    },
    {
      "date": "2026-10-22",
      "company": "Hugging Face",
      "model": "Carbon-X",
      "kind": "Science",
      "summary": "An open model that enables conversational queries about DNA sequences and their effects on gene activity and health.",
      "availability": "Open model that runs locally on a laptop without sending data to Hugging Face",
      "source_name": "Wired AI",
      "source_url": "https://www.wired.com/story/hugging-face-new-ai-biology-model-chat-with-dna/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-carbonx"
    },
    {
      "date": "2026-10-20",
      "company": "Mistral",
      "model": "Voxtral 3",
      "kind": "Audio",
      "summary": "Speech-to-text and text-to-speech models with voice agent SDK supporting 50+ languages for real-time conversations.",
      "availability": "Available via Mistral's EU-hosted API or for self-deployment on own infrastructure",
      "source_name": "Mistral AI",
      "source_url": "https://mistral.ai/news/voxtral-3/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-voxtral3"
    },
    {
      "date": "2026-10-20",
      "company": "Black Forest Labs",
      "model": "FLUX.3 Image Klein",
      "kind": "Image",
      "summary": "Open image generation model supporting text-to-image, image-to-image, and bounding box layout control.",
      "availability": "Open checkpoints available, runs in Diffusers, accessible via Hugging Face space",
      "source_name": "HuggingFace Blog",
      "source_url": "https://huggingface.co/blog/flux-3",
      "tracker_url": "https://www.aidaynews.com/tracker#m-flux3imageklein"
    },
    {
      "date": "2026-10-18",
      "company": "IBM",
      "model": "Granite Switch 4.1",
      "kind": "Text",
      "summary": "Modular models combining chat, decision-making, and speech-to-text capabilities in one endpoint.",
      "availability": "Available in 8B and 30B sizes, runs in vLLM, llama.cpp, and downstream platforms.",
      "source_name": "HuggingFace Blog",
      "source_url": "https://huggingface.co/blog/ibm-granite/granite-switch-4-1",
      "tracker_url": "https://www.aidaynews.com/tracker#m-graniteswitch41"
    },
    {
      "date": "2026-10-17",
      "company": "Google DeepMind",
      "model": "Gemini 4 Argon",
      "kind": "Robotics",
      "summary": "A physical reasoning model for robotics that scores 87% on the Embodied Reasoning benchmark, outperforming other frontier models.",
      "availability": "",
      "source_name": "Google DeepMind",
      "source_url": "https://deepmind.google/blog/gemini-4-argon-for-robotics/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-gemini4argon"
    },
    {
      "date": "2026-10-17",
      "company": "Reflection",
      "model": "Beam",
      "kind": "Text",
      "summary": "501B parameter mixture-of-experts model with 23B active parameters per token, supporting 1M token context window.",
      "availability": "Open source under Apache 2.0 license on Hugging Face",
      "source_name": "HuggingFace Blog",
      "source_url": "https://huggingface.co/blog/reflection-beam",
      "tracker_url": "https://www.aidaynews.com/tracker#m-beam"
    },
    {
      "date": "2026-10-15",
      "company": "Liquid AI",
      "model": "LFM2.5-24B-A2B",
      "kind": "Text",
      "summary": "A 24B parameter reasoning model with 2B active parameters designed for fast on-device agents.",
      "availability": "Open-weight model available on Hugging Face",
      "source_name": "HuggingFace Blog",
      "source_url": "https://huggingface.co/blog/LiquidAI/lfm25-24b-a2b",
      "tracker_url": "https://www.aidaynews.com/tracker#m-lfm2524ba2b"
    },
    {
      "date": "2026-10-10",
      "company": "Technology Innovation Institute",
      "model": "Falcon-ASR",
      "kind": "Audio",
      "summary": "1.6 billion parameter speech recognition model for Arabic with focus on Emirati dialect, also supporting English, French, Spanish and Portuguese.",
      "availability": "Hugging Face Demo Space available; API access and native applications planned",
      "source_name": "HuggingFace Blog",
      "source_url": "https://huggingface.co/blog/tiiuae/falcon-asr",
      "tracker_url": "https://www.aidaynews.com/tracker#m-falconasr"
    },
    {
      "date": "2026-10-07",
      "company": "Google DeepMind",
      "model": "EmbeddingGemma 2",
      "kind": "Other",
      "summary": "Multimodal embedding model mapping text, images, audio, and video into unified embedding space with 740 million parameters.",
      "availability": "Apache 2.0 license, available on Hugging Face and Kaggle",
      "source_name": "Google DeepMind",
      "source_url": "https://deepmind.google/blog/embeddinggemma-2-an-open-lightweight-multimodal-embedding-model/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-embeddinggemma2"
    },
    {
      "date": "2026-10-06",
      "company": "Technology Innovation Institute",
      "model": "Falcon-Emirati-7B",
      "kind": "Text",
      "summary": "A dialect-specialized model built to understand and generate Emirati Arabic with cultural context and nuance.",
      "availability": "",
      "source_name": "HuggingFace Blog",
      "source_url": "https://huggingface.co/blog/tiiuae/falcon-emirati",
      "tracker_url": "https://www.aidaynews.com/tracker#m-falconemirati7b"
    },
    {
      "date": "2026-09-26",
      "company": "Google DeepMind",
      "model": "Gemini 3.8 Flash Cyber",
      "kind": "Code",
      "summary": "Specialized cyber defense model that autonomously finds and fixes vulnerabilities in code at reduced operating cost.",
      "availability": "Limited access through Fairwind Program for governments, critical infrastructure operators, and trusted enterprise partners",
      "source_name": "Google DeepMind",
      "source_url": "https://deepmind.google/blog/proactive-cyber-defense-for-governments-and-enterprises/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-gemini38flashcyber"
    },
    {
      "date": "2026-09-25",
      "company": "Liquid AI",
      "model": "LFM2.5-VL-DSpark",
      "kind": "Other",
      "summary": "Experimental DSpark draft model for vision-language model LFM2.5-VL-3B enabling faster inference through speculative decoding.",
      "availability": "Open-weight. Available on Hugging Face in Safetensors and GGUF formats.",
      "source_name": "HuggingFace Blog",
      "source_url": "https://huggingface.co/blog/LiquidAI/lfm2-5-vl-dspark",
      "tracker_url": "https://www.aidaynews.com/tracker#m-lfm25vldspark"
    },
    {
      "date": "2026-09-19",
      "company": "TypeSafe AI",
      "model": "Jev",
      "kind": "Other",
      "summary": "A transformer-based model that produces calibrated decisions instead of text, designed for software automation tasks.",
      "availability": "Available via API; developers can access it through TypeSafe AI's service.",
      "source_name": "TechCrunch AI",
      "source_url": "https://techcrunch.com/2026/09/18/a-new-kind-of-ai-model-from-a-chatgpt-inventor-is-thrilling-developers/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-jev"
    },
    {
      "date": "2026-09-16",
      "company": "Google DeepMind",
      "model": "Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking",
      "kind": "Audio",
      "summary": "Two advanced voice dialogue models with improved reasoning, real-time visual processing, and background task execution capabilities.",
      "availability": "Available today via Gemini API, Google Workspace, and Gemini app",
      "source_name": "Google DeepMind",
      "source_url": "https://deepmind.google/blog/introducing-gemini-3-8-live-and-3-8-live-extended-thinking/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-gemini38liveandgemini38liveextendedthinking"
    },
    {
      "date": "2026-09-15",
      "company": "Salesforce",
      "model": "Koa",
      "kind": "Text",
      "summary": "Reasoning model built on Nvidia's Nemotron, trained for sales, marketing, and customer support tasks.",
      "availability": "Provided as alternative in Salesforce's Agentforce platform for enterprise customers.",
      "source_name": "TechCrunch AI",
      "source_url": "https://techcrunch.com/2026/09/15/salesforce-and-nvidias-new-reasoning-model-is-everything-the-ai-labs-should-fear/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-koa"
    },
    {
      "date": "2026-09-09",
      "company": "Suno",
      "model": "Suno v6",
      "kind": "Audio",
      "summary": "New model family trained on licensed music data from major music labels and distributors.",
      "availability": "Base and wild versions for paying users, mini version for all users",
      "source_name": "TechCrunch AI",
      "source_url": "https://techcrunch.com/2026/09/09/suno-replaces-its-ai-models-with-a-new-one-trained-on-licensed-music-as-copyright-suits-pile-up/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-v6"
    },
    {
      "date": "2026-09-08",
      "company": "IBM",
      "model": "Granite Time Series PatchTST-FM-r2",
      "kind": "Other",
      "summary": "Time series forecasting model with conformer architecture, larger pretraining corpus, probabilistic forecasting, and missing value imputation.",
      "availability": "Open weights and code available under Apache 2.0 and OpenMDW 1.0 licenses",
      "source_name": "HuggingFace Blog",
      "source_url": "https://huggingface.co/blog/ibm-research/ibm-releases-sota-granite-time-series",
      "tracker_url": "https://www.aidaynews.com/tracker#m-granitetimeseriespatchtstfmr2"
    },
    {
      "date": "2026-09-04",
      "company": "Google DeepMind",
      "model": "WeatherNext 3",
      "kind": "Science",
      "summary": "AI weather model using real-time satellite data for hourly high-resolution forecasts five times sharper than previous version.",
      "availability": "Integrated into Google Search, Gemini, Maps, and Google Cloud Platform.",
      "source_name": "Google DeepMind",
      "source_url": "https://deepmind.google/blog/introducing-weathernext-3-our-most-advanced-and-accurate-global-weather-ai-model/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-weathernext3"
    },
    {
      "date": "2026-09-04",
      "company": "OpenAI",
      "model": "GPT-6 Astra",
      "kind": "Other",
      "summary": "Navigates computers and web browsers, writes software, and solves difficult math problems.",
      "availability": "Paid customers starting with Daybreak early access program, then ChatGPT Plus, Pro, Business and Enterprise subscribers",
      "source_name": "Wired AI",
      "source_url": "https://www.wired.com/story/openai-says-gpt-6-can-use-a-computer-better-than-a-human/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-gpt6astra"
    },
    {
      "date": "2026-09-03",
      "company": "Google DeepMind",
      "model": "Gemini 3.8 Flash",
      "kind": "Text",
      "summary": "Google released Gemini 3.8 Flash with improved reasoning and coding capabilities, available in standard and Cyber variants.",
      "availability": "Standard version available across Google ecosystem; Cyber version limited to trusted testers and governments",
      "source_name": "Ars Technica AI",
      "source_url": "https://arstechnica.com/ai/2026/09/google-releases-gemini-3-8-flash-its-third-flash-model-in-six-weeks/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-gemini38flash"
    },
    {
      "date": "2026-08-27",
      "company": "IBM",
      "model": "Granite 4.2",
      "kind": "Text",
      "summary": "Open-weight models in 3B, 8B, and 30B variants with 128,000-token context window and reasoning capabilities.",
      "availability": "Open-weight models designed to be downloaded and self-hosted.",
      "source_name": "Ars Technica AI",
      "source_url": "https://arstechnica.com/ai/2026/08/ibms-new-granite-4-2-models-ride-the-wave-of-interest-in-local-llms/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-granite42"
    },
    {
      "date": "2026-08-27",
      "company": "Google DeepMind",
      "model": "Gemini Omni 1.1 Flash",
      "kind": "Video",
      "summary": "Production-ready update enabling developers to extend scenes, control frame transitions, and generate 4K video with improved visual consistency.",
      "availability": "Available via Google AI Studio and Gemini API for developers",
      "source_name": "Google DeepMind",
      "source_url": "https://deepmind.google/blog/gemini-omni-1-1-flash-lets-you-build-with-more-control/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-geminiomni11flash"
    },
    {
      "date": "2026-08-27",
      "company": "Google DeepMind",
      "model": "Gemini 3.5 Transcribe",
      "kind": "Audio",
      "summary": "Speech-to-text model for real-time transcription with improved word error rates, multi-speaker identification, and support for over 85 languages.",
      "availability": "Available via Gemini API, Google AI Studio, Gemini Enterprise Agent Platform, and integrated into Gboard, Gemini app, and Chrome.",
      "source_name": "Google DeepMind",
      "source_url": "https://deepmind.google/blog/intelligent-transcription-with-gemini-3-5-transcribe/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-gemini35transcribe"
    },
    {
      "date": "2026-08-23",
      "company": "Z.ai",
      "model": "GLM-5.3",
      "kind": "Text",
      "summary": "Extended post-training version of GLM-5.2 base model with improved benchmark performance on coding tasks.",
      "availability": "Currently in coding plan, coming to API and Hugging Face open weights in two weeks",
      "source_name": "Interconnects",
      "source_url": "https://www.interconnects.ai/p/glm-53-how-chinese-labs-keep-stride",
      "tracker_url": "https://www.aidaynews.com/tracker#m-glm53"
    },
    {
      "date": "2026-08-21",
      "company": "NVIDIA",
      "model": "Magpie Multilingual TTS",
      "kind": "Audio",
      "summary": "Multilingual TTS model expanded to 12 languages including Modern Standard Arabic, Korean, and Brazilian Portuguese with improved quality.",
      "availability": "Open weights on Hugging Face and production-ready NVIDIA NIM for own infrastructure deployment.",
      "source_name": "HuggingFace Blog",
      "source_url": "https://huggingface.co/blog/nvidia/magpie-tts-multilingual-voice-agents",
      "tracker_url": "https://www.aidaynews.com/tracker#m-magpiemultilingualtts"
    },
    {
      "date": "2026-08-16",
      "company": "Writer",
      "model": "Palmyra X6",
      "kind": "Text",
      "summary": "Post-training variation on GLM-5.2 designed to reduce deployment costs by up to 50% for basic tasks.",
      "availability": "Available to Writer clients starting Thursday",
      "source_name": "TechCrunch AI",
      "source_url": "https://techcrunch.com/2026/08/13/writer-introduces-new-ai-model-and-upgraded-harness-to-contain-token-costs/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-palmyrax6"
    },
    {
      "date": "2026-08-15",
      "company": "Liquid AI",
      "model": "LFM2.5-VL-3B",
      "kind": "Image",
      "summary": "Vision-language model with improved screen understanding, grounding, multi-image reasoning, and function calling for edge devices.",
      "availability": "Available on Hugging Face. Can run on-device or via GPU inference.",
      "source_name": "HuggingFace Blog",
      "source_url": "https://huggingface.co/blog/LiquidAI/lfm2-5-vl-3b",
      "tracker_url": "https://www.aidaynews.com/tracker#m-lfm25vl3b"
    },
    {
      "date": "2026-08-15",
      "company": "Google DeepMind",
      "model": "Gemini Robotics ER 2",
      "kind": "Robotics",
      "summary": "A robotics model enabling real-time spatial reasoning, multi-step task planning, and multi-robot collaboration with video understanding.",
      "availability": "Publicly available via Gemini API, Google AI Studio, and Gemini Enterprise Agent Platform private preview",
      "source_name": "Google DeepMind",
      "source_url": "https://deepmind.google/blog/gemini-robotics-er-2-powering-robotics-with-video-understanding-task-orchestration-and-multi-robot-collaboration/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-geminiroboticser2"
    },
    {
      "date": "2026-08-14",
      "company": "Google DeepMind",
      "model": "Gemini 3.7 Flash",
      "kind": "Code",
      "summary": "Improved coding and agent model with better performance on debugging, code generation, and web development tasks.",
      "availability": "Gemini API, Google AI Studio, Android Studio, Gemini Enterprise, Spark for Pro and Ultra subscribers",
      "source_name": "Google DeepMind",
      "source_url": "https://deepmind.google/blog/introducing-gemini-3-7-flash/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-gemini37flash"
    },
    {
      "date": "2026-08-11",
      "company": "Meta",
      "model": "Muse Glimmer",
      "kind": "Other",
      "summary": "Multimodal 30B parameter model distilled from Muse, designed for local agentic use cases with vision and language capabilities.",
      "availability": "Open source under Apache 2.0 license, available on Hugging Face Hub for local deployment.",
      "source_name": "HuggingFace Blog",
      "source_url": "https://huggingface.co/blog/muse-glimmer",
      "tracker_url": "https://www.aidaynews.com/tracker#m-museglimmer"
    },
    {
      "date": "2026-08-08",
      "company": "Moonshot AI",
      "model": "Kimi K3",
      "kind": "Text",
      "summary": "Chinese language model released with large context window and improved capabilities across multiple domains.",
      "availability": "Available via API with subscription plans of $40, $100, and $200 per month",
      "source_name": "Interconnects",
      "source_url": "https://www.interconnects.ai/p/open-models-recap-more-on-kimi-k3",
      "tracker_url": "https://www.aidaynews.com/tracker#m-kimik3"
    },
    {
      "date": "2026-08-05",
      "company": "Liquid AI",
      "model": "LFM2.5-2.6B",
      "kind": "Text",
      "summary": "A 2.6 billion parameter model designed for on-device agent deployment with tool calling and multi-step workflows.",
      "availability": "Available on Hugging Face for download and use",
      "source_name": "HuggingFace Blog",
      "source_url": "https://huggingface.co/blog/LiquidAI/lfm2-5-2-6b",
      "tracker_url": "https://www.aidaynews.com/tracker#m-lfm2526b"
    },
    {
      "date": "2026-08-02",
      "company": "AI2",
      "model": "OlmoEarth",
      "kind": "Science",
      "summary": "Family of Earth observation foundation models pretrained on roughly 10 terabytes of multimodal satellite data.",
      "availability": "Governments, NGOs, and mission-driven organizations can adapt OlmoEarth for applications.",
      "source_name": "HuggingFace Blog",
      "source_url": "https://huggingface.co/blog/allenai/olmoearth-infrastructure",
      "tracker_url": "https://www.aidaynews.com/tracker#m-olmoearth"
    },
    {
      "date": "2026-08-02",
      "company": "Liquid AI",
      "model": "LFM2.5-Encoder-230M and LFM2.5-Encoder-350M",
      "kind": "Text",
      "summary": "Two encoder models matching larger models' quality while supporting 8,192-token context with fast CPU inference.",
      "availability": "Available on Hugging Face for download and fine-tuning.",
      "source_name": "HuggingFace Blog",
      "source_url": "https://huggingface.co/blog/LiquidAI/lfm2-5-encoders",
      "tracker_url": "https://www.aidaynews.com/tracker#m-lfm25encoder230mandlfm25encoder350m"
    },
    {
      "date": "2026-07-31",
      "company": "Google DeepMind",
      "model": "Gemini Robotics 2",
      "kind": "Robotics",
      "summary": "AI model combining vision language and action models to control robots performing complex physical tasks.",
      "availability": "",
      "source_name": "Wired AI",
      "source_url": "https://www.wired.com/story/google-gemini-can-control-humanoid-robots/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-geminirobotics2"
    },
    {
      "date": "2026-07-30",
      "company": "Google DeepMind",
      "model": "Lyria 3.5",
      "kind": "Audio",
      "summary": "Music generation model with improved musicality, lyrics, vocals, and creative control features.",
      "availability": "Rolling out today in Google Flow Music",
      "source_name": "Google DeepMind",
      "source_url": "https://deepmind.google/blog/were-launching-lyria-35-in-google-flow-music-with-advances-across-musicality-lyrics-vocals-and-creative-control/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-lyria35"
    },
    {
      "date": "2026-07-27",
      "company": "NVIDIA",
      "model": "Cosmos-H-Dreams",
      "kind": "Robotics",
      "summary": "Real-time action-conditioned generative simulator for surgical robotics that distills Cosmos-H-Surgical-Simulator for interactive closed-loop control.",
      "availability": "Runs on single NVIDIA RTX PRO 6000 GPU; integrated with dVRK and Versius platforms.",
      "source_name": "HuggingFace Blog",
      "source_url": "https://huggingface.co/blog/nvidia/cosmos-h-dreams",
      "tracker_url": "https://www.aidaynews.com/tracker#m-cosmoshdreams"
    },
    {
      "date": "2026-07-25",
      "company": "OpenAI",
      "model": "ChatGPT-Live",
      "kind": "Audio",
      "summary": "New family of voice models enabling ChatGPT Voice to control AI agents and perform tasks on computers.",
      "availability": "Rolling out globally in ChatGPT desktop app and iOS app via remote access",
      "source_name": "TechCrunch AI",
      "source_url": "https://techcrunch.com/2026/07/24/openais-new-voice-mode-makes-it-to-the-chatgpt-desktop-app/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-chatgptlive"
    },
    {
      "date": "2026-07-24",
      "company": "OpenAI",
      "model": "GPT 5.6-Luna",
      "kind": "Text",
      "summary": "Smallest model from latest release outperforms GPT 5.5 on HealthBench evaluation for health queries.",
      "availability": "Logged-in U.S. users with free, Go, Plus, Pro plans on web and iOS",
      "source_name": "TechCrunch AI",
      "source_url": "https://techcrunch.com/2026/07/23/openai-makes-chatgpt-health-available-to-all-u-s-users/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-gpt56luna"
    },
    {
      "date": "2026-07-22",
      "company": "Google DeepMind",
      "model": "Gemini 3.6 Flash",
      "kind": "Text",
      "summary": "More efficient than 3.5 Flash with better coding, knowledge work, and multimodal performance at lower cost.",
      "availability": "Available via Gemini API and Gemini Enterprise",
      "source_name": "Google DeepMind",
      "source_url": "https://deepmind.google/blog/introducing-gemini-36-flash-35-flash-lite-and-35-flash-cyber/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-gemini36flash"
    },
    {
      "date": "2026-07-22",
      "company": "Google DeepMind",
      "model": "Gemini 3.5 Flash Cyber",
      "kind": "Code",
      "summary": "Lightweight cybersecurity model fine-tuned to find, validate, and patch vulnerabilities quickly and efficiently.",
      "availability": "Limited-access pilot via CodeMender, exclusively for governments and trusted partners.",
      "source_name": "Google DeepMind",
      "source_url": "https://deepmind.google/blog/introducing-gemini-3-5-flash-cyber/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-gemini35flashcyber"
    },
    {
      "date": "2026-07-21",
      "company": "NVIDIA",
      "model": "Cosmos 3 Edge",
      "kind": "Robotics",
      "summary": "A 4-billion-parameter open world model for robots and vision AI agents to understand surroundings and generate actions on edge devices.",
      "availability": "Open model available on Hugging Face for download and deployment on NVIDIA edge devices.",
      "source_name": "HuggingFace Blog",
      "source_url": "https://huggingface.co/blog/nvidia/cosmos3edge",
      "tracker_url": "https://www.aidaynews.com/tracker#m-cosmos3edge"
    },
    {
      "date": "2026-07-19",
      "company": "Thinking Machines",
      "model": "Inkling 2 Small",
      "kind": "Image",
      "summary": "Second generation multimodal model with 594B total and 17B active parameters accepting text and image inputs.",
      "availability": "Open model available on Hugging Face Hub",
      "source_name": "HuggingFace Blog",
      "source_url": "https://huggingface.co/blog/thinkingmachines-inkling",
      "tracker_url": "https://www.aidaynews.com/tracker#m-inkling2small"
    },
    {
      "date": "2026-07-16",
      "company": "Applied Computing",
      "model": "Orbital",
      "kind": "Other",
      "summary": "Foundation AI model combining time series, physics-based, and language models to predict oil and gas facility states.",
      "availability": "In use at large publicly listed upstream oil and gas, downstream refining and petrochemicals companies",
      "source_name": "TechCrunch AI",
      "source_url": "https://techcrunch.com/2026/07/15/applied-computing-wants-to-give-oil-and-gas-operators-an-ai-model-for-the-entire-plant/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-orbital"
    },
    {
      "date": "2026-07-16",
      "company": "Thinking Machines Lab",
      "model": "Inkling",
      "kind": "Text",
      "summary": "Open-weight mixture-of-experts model with 975B parameters trained on multimodal data, outputs text and code.",
      "availability": "Open-weight, downloadable and modifiable by outside developers and companies.",
      "source_name": "TechCrunch AI",
      "source_url": "https://techcrunch.com/2026/07/15/thinking-machines-amps-up-its-bet-against-one-size-fits-all-ai-with-its-first-open-model-inkling/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-inkling"
    },
    {
      "date": "2026-07-12",
      "company": "Meta",
      "model": "Muse Spark 1.1",
      "kind": "Text",
      "summary": "Multimodal reasoning model with improvements in tool use, coding, and computer understanding for agentic tasks.",
      "availability": "Public preview via Meta Model API and Meta AI app at meta.ai",
      "source_name": "AI at Meta",
      "source_url": "https://ai.meta.com/blog/introducing-muse-spark-meta-model-api/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-musespark11"
    },
    {
      "date": "2026-07-10",
      "company": "Meta",
      "model": "Muse Spark",
      "kind": "Text",
      "summary": "New AI model better at coding, performing at or near levels of leading competitors, with paid version option.",
      "availability": "Paid version available to customers at lower cost than competing models",
      "source_name": "The New York Times",
      "source_url": "https://www.nytimes.com/2026/07/09/technology/meta-muse-spark-artificial-intelligence.html",
      "tracker_url": "https://www.aidaynews.com/tracker#m-musespark"
    },
    {
      "date": "2026-07-10",
      "company": "OpenAI",
      "model": "GPT-5.6 Sol",
      "kind": "Text",
      "summary": "OpenAI released GPT-5.6 Sol, its flagship artificial intelligence model with capabilities for office work and cybersecurity tasks.",
      "availability": "Publicly released to individuals and businesses",
      "source_name": "The New York Times",
      "source_url": "https://www.nytimes.com/2026/07/09/technology/openai-sol-ai.html",
      "tracker_url": "https://www.aidaynews.com/tracker#m-gpt56sol"
    },
    {
      "date": "2026-07-10",
      "company": "OpenAI",
      "model": "GPT-5.6",
      "kind": "Text",
      "summary": "New model family with three variants offering improved efficiency, cost-effectiveness, and cybersecurity capabilities.",
      "availability": "Available across ChatGPT, Codex, and the OpenAI API",
      "source_name": "TechCrunch AI",
      "source_url": "https://techcrunch.com/2026/07/09/openai-launches-its-new-family-of-models-with-gpt-5-6/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-gpt56"
    },
    {
      "date": "2026-07-09",
      "company": "SpaceXAI",
      "model": "Grok 4.5",
      "kind": "Text",
      "summary": "SpaceXAI released Grok 4.5, a model with twice greater token efficiency than competitors, designed for coding, writing, research, and knowledge work.",
      "availability": "Available to the public, with beta testing completed",
      "source_name": "TechCrunch AI",
      "source_url": "https://techcrunch.com/2026/07/08/spacexai-releases-grok-4-5-which-elon-describes-as-an-opus-class-model/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-grok45"
    },
    {
      "date": "2026-07-09",
      "company": "OpenAI",
      "model": "GPT-Live-1 and GPT-Live-1 mini",
      "kind": "Audio",
      "summary": "Full-duplex voice models that sound more natural, handle turn-taking better, and allow users to interrupt naturally in conversations.",
      "availability": "GPT-Live-1 mini replaces Advanced Voice Mode by default; GPT-Live-1 for paid tier users.",
      "source_name": "TechCrunch AI",
      "source_url": "https://techcrunch.com/2026/07/08/openai-releases-new-voice-models-for-more-natural-live-conversations/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-gptlive1andgptlive1mini"
    },
    {
      "date": "2026-07-01",
      "company": "Google DeepMind",
      "model": "Nano Banana 2 Lite",
      "kind": "Image",
      "summary": "Fast and cheap image generation model that creates images in about 4 seconds at $0.034 per 1K images.",
      "availability": "Google AI Studio, API, and Gemini by selecting Flash-Lite option",
      "source_name": "Ars Technica AI",
      "source_url": "https://arstechnica.com/ai/2026/06/googles-new-nano-banana-2-lite-image-model-is-its-fastest-and-cheapest-yet/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-nanobanana2lite"
    },
    {
      "date": "2026-07-01",
      "company": "Anthropic",
      "model": "Claude Sonnet 5",
      "kind": "Text",
      "summary": "More powerful agentic midsize model with improved reasoning, tool use, coding, and knowledge work capabilities.",
      "availability": "Default for free and Pro plans, available for every subscription starting Tuesday",
      "source_name": "TechCrunch AI",
      "source_url": "https://techcrunch.com/2026/06/30/anthropic-launches-claude-sonnet-5-as-a-cheaper-way-to-run-agents/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-claudesonnet5"
    },
    {
      "date": "2026-06-30",
      "company": "Base44",
      "model": "Base1",
      "kind": "Code",
      "summary": "A custom LLM trained on tens of millions of real user interactions to support app creation with natural language.",
      "availability": "Rolling out to Base44 users",
      "source_name": "TechCrunch AI",
      "source_url": "https://techcrunch.com/2026/06/29/vibe-coding-platform-base44-launches-own-model-as-ai-startups-seek-defensibility/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-base1"
    },
    {
      "date": "2026-06-25",
      "company": "ByteDance",
      "model": "Seedance 2.5",
      "kind": "Video",
      "summary": "Generates 30-second 4K videos from a single prompt with up to 50 reference pieces for greater control.",
      "availability": "Expected to launch in China next month; may be available July 9.",
      "source_name": "CNET",
      "source_url": "https://www.cnet.com/tech/services-and-software/bytedance-introduces-new-seedance-2-5-video-model/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-seedance25"
    },
    {
      "date": "2026-06-21",
      "company": "Zhipu AI",
      "model": "GLM-5.2",
      "kind": "Code",
      "summary": "Long-context model with 1M token context, improved coding capabilities, and IndexShare architecture optimization.",
      "availability": "MIT open-source license, no regional limits",
      "source_name": "HuggingFace Blog",
      "source_url": "https://huggingface.co/blog/zai-org/glm-52-blog",
      "tracker_url": "https://www.aidaynews.com/tracker#m-glm52"
    },
    {
      "date": "2026-06-17",
      "company": "Google DeepMind",
      "model": "Lyria 3",
      "kind": "Audio",
      "summary": "Music generation model that creates music tracks from text prompts and images.",
      "availability": "Available in Gemini app on Android 17 Pixel devices",
      "source_name": "TechCrunch AI",
      "source_url": "https://techcrunch.com/2026/06/16/android-17-launches-with-new-multitasking-tools-as-google-expands-gemini-features/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-lyria3"
    },
    {
      "date": "2026-06-11",
      "company": "Google DeepMind",
      "model": "DiffusionGemma",
      "kind": "Text",
      "summary": "A text generation model that produces outputs in parallel rather than sequentially, achieving 4x faster speeds on local hardware.",
      "availability": "Available under Apache 2.0 license, downloadable from Hugging Face.",
      "source_name": "Ars Technica AI",
      "source_url": "https://arstechnica.com/google/2026/06/googles-latest-diffusiongemma-open-ai-model-comes-with-a-4x-speed-boost/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-diffusiongemma"
    },
    {
      "date": "2026-06-10",
      "company": "Google DeepMind",
      "model": "Gemini 3.5 Live Translate",
      "kind": "Audio",
      "summary": "Audio model delivering near real-time speech-to-speech translation in over 70 languages with natural intonation and continuous generation.",
      "availability": "Developers via Gemini Live API, enterprises in Google Meet, everyone via Google Translate app.",
      "source_name": "Google DeepMind",
      "source_url": "https://deepmind.google/blog/fluid-natural-voice-translation-with-gemini-35-live-translate/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-gemini35livetranslate"
    },
    {
      "date": "2026-06-10",
      "company": "Anthropic",
      "model": "Claude Fable 5",
      "kind": "Text",
      "summary": "First publicly available version of Mythos model with safety limits on high-risk areas.",
      "availability": "Claude API, Enterprise plans, with staged rollout to subscriptions through June 22.",
      "source_name": "TechCrunch AI",
      "source_url": "https://techcrunch.com/2026/06/09/anthropics-claude-fable-5-is-a-version-of-mythos-the-public-can-access-today/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-claudefable5"
    },
    {
      "date": "2026-06-05",
      "company": "Google DeepMind",
      "model": "Gemma 4 12B",
      "kind": "Text",
      "summary": "A 12-billion-parameter model optimized to run on consumer laptops with 16GB of RAM.",
      "availability": "Available for download on Kaggle and Hugging Face. Accessible via LM Studio and Google AI Edge Gallery.",
      "source_name": "Ars Technica AI",
      "source_url": "https://arstechnica.com/google/2026/06/googles-new-gemma-4-open-ai-model-is-sized-for-your-laptop/",
      "tracker_url": "https://www.aidaynews.com/tracker#m-gemma412b"
    },
    {
      "date": "2026-06-03",
      "company": "Microsoft",
      "model": "MAI-Code-1-Flash",
      "kind": "Code",
      "summary": "AI coding model that converts written descriptions into source code for applications and websites.",
      "availability": "Available in GitHub Copilot and Visual Studio Code text editor",
      "source_name": "CNBC",
      "source_url": "https://www.cnbc.com/2026/06/02/microsoft-unveils-new-ai-models-lessen-reliance-on-openai-lower-costs.html",
      "tracker_url": "https://www.aidaynews.com/tracker#m-maicode1flash"
    }
  ]
}