EVENTS76
ARTICLES52
CATEGORYAI
Active
New AI model releases, benchmarks, and capabilities. Foundation models, open-source releases, and API launches from leading AI labs.
LATEST EVENTS
| Entity | Type | Summary | Impact | Confidence |
|---|---|---|---|---|
| Arcee | product | Arcee released AI Trinity-Large-Thinking, a 399-billion parameter open-source text-only reasoning model for customizabil… | medium | 100% |
| llama.cpp | product | TurboQuant model weight compression support was added to a fork of llama.cpp, though its integration into the main proje… | low | 80% |
| Apple | announcement | Apple is strategically building an underlying orchestration layer for AI agents to integrate various AI models into its … | high | 90% |
| Databricks | product | Databricks announced the general availability and open sourcing of Unity Catalog Business Semantics for defining busines… | medium | 90% |
| Mistral | product | Mistral unveiled Forge, a platform enabling enterprises to build custom AI models from scratch. | medium | 100% |
| OpenAI | product | OpenAI released GPT-5.4 mini and nano, smaller models optimized for high-volume, faster, and lower-cost workloads. | high | 100% |
| Anthropic | product | Anthropic introduced 'Claude Mythos,' a new tier of models with dramatically higher scores in coding, reasoning, and cyb… | high | 100% |
| Meta, Google | partnership | Meta is potentially exploring licensing Google's Gemini technology to route some AI requests through Gemini models. | medium | 90% |
| ChatGPT (OpenAI) | regulatory | Check Point research revealed a vulnerability in ChatGPT allowing sensitive data exfiltration without user alerts. | high | 100% |
| product | Google launched Gemma 4, a new family of open-weight AI models, described as its 'most intelligent' to date, under an Ap… | high | 100% | |
| PrismML | announcement | PrismML potentially achieved a breakthrough in local LLM inference with lossless 1-bit quantization. | medium | 80% |
| Microsoft | product | Microsoft released three new AI models. | medium | 80% |
| Anthropic | regulatory | Anthropic experienced a Claude Code source code leak, revealing unreleased models like Opus 4.7, Sonnet 4.8, and 'Mythos… | high | 90% |
| OpenAI | financial | OpenAI secured a $122 billion funding round. | high | 90% |
| product | Google released Gemma 4, a new family of open-weight AI models, as an open-source competitor to Llama. | high | 90% | |
| Positronic Robotics | product | Positronic Robotics launched its 'Physical AI Leaderboard' (PhAIL) to benchmark robotics foundation models on real hardw… | medium | 100% |
| product | Google rolled out updates to its Gemini AI, expanding voice search and 'Personal Intelligence,' and releasing Gemini 3.1… | high | 100% | |
| product | Google's Gemini Live received an upgrade enabling more detailed and interactive news summaries, allowing users to ask fo… | medium | 90% | |
| llama.cpp | product | A bug in llama.cpp that silently disabled Vulkan GPU on 32-bit ARM devices, causing a performance bottleneck, has been f… | medium | 90% |
| Ollama, Apple | product | Ollama 0.19 preview integrates Apple's MLX framework, significantly boosting local Llama model inference speed on Mac de… | medium | 100% |
| OpenAI | announcement | A YouTube video speculates GPT-5 will be an orchestrated system of specialized AI agents, focusing on autonomous capabil… | low | 70% |
| US Lawmakers, FTC | regulatory | A bipartisan House bill proposes tasking the FTC with establishing transparency requirements for AI foundation models, e… | high | 90% |
| OpenAI | financial | OpenAI closed a $122 billion funding round, valuing the company at $852 billion, to accelerate frontier AI research and … | high | 100% |
| OpenAI | product | OpenAI patched a critical ChatGPT data exfiltration flaw and a Codex GitHub token vulnerability, enhancing AI system sec… | high | 100% |
| UCL School of Management, Professor Angela Aristidou | announcement | Professor Angela Aristidou argues current AI benchmarks are flawed, advocating for evaluating AI performance within real… | low | 80% |
| OpenAI, Anthropic, Google, Meta | announcement | The LLM ecosystem is rapidly evolving with over 500 models available, offering developers unprecedented choice and ongoi… | low | 80% |
| Google DeepMind | product | Google DeepMind launched Gemma 4, open-weight AI models designed to run on a single GPU with strong benchmarks, competin… | high | 100% |
| NVIDIA | product | NVIDIA's Blackwell Ultra GPUs set new MLPerf Inference v6.0 records, demonstrating highest throughput across various AI … | high | 100% |
| Anthropic | product | The Claude Code leak revealed Anthropic tracks users' vulgar language for analytics, raising privacy concerns about inte… | medium | 90% |
| Anthropic | product | Anthropic blocked Claude Pro/Max subscribers from using third-party AI agent frameworks like OpenClaw, shifting costs to… | high | 100% |
| Anthropic, xAI, OpenAI, Google, Mistral | announcement | April 2026 is active for AI model releases, with Anthropic testing Claude Mythos, xAI targeting Grok 5, and OpenAI itera… | low | 80% |
| Linux Foundation, Coinbase | partnership | The Linux Foundation launched the x402 Foundation, a neutral home for Coinbase's x402 protocol, to embed payments into w… | medium | 80% |
| MLCommons | announcement | MLCommons released new MLPerf Inference v6.0 benchmark results, including an open-weight LLM benchmark based on GPT-OSS … | medium | 90% |
| Anthropic | product | A critical vulnerability was discovered in Anthropic's Claude Code following its source code leak, potentially allowing … | high | 90% |
| Anthropic | product | Anthropic accidentally released Claude Code's internal source code due to human error, raising security questions and be… | high | 100% |
| OpenAI | product | OpenAI's GPT-5.4 Pro achieved an IQ score of 150 on the Mensa Norway test, setting new state-of-the-art records. | high | 90% |
| Anthropic | product | Anthropic confirmed its Claude Code AI assistant's internal source code was accidentally leaked via an npm packaging err… | high | 100% |
| Meta | announcement | An internal leak suggests Meta is evaluating new open-source "Avocado" models, including multimodal agentic variants. | medium | 90% |
| product | Google's Gemini powers Siri's cloud reasoning on iPhone and the native assistant on Android, securing extensive mobile d… | high | 90% | |
| Apple | announcement | Apple is building an orchestration layer for AI agents using App Intents and MCP, allowing various models like ChatGPT a… | high | 90% |
| Databricks | product | Databricks announced the General Availability and open-sourcing of Unity Catalog Business Semantics in Apache Spark. | medium | 100% |
| Meta | product | Meta's Llama models are struggling with complex math problems, leading some Meta AI requests to be routed through Google… | medium | 100% |
| Anthropic | product | Anthropic announced general availability of 1M context for Claude Opus 4.6 and Sonnet 4.6 models. | medium | 100% |
| OpenAI | product | OpenAI released GPT-5.4 Mini and Nano models, designed for high-volume, faster, and lower-cost AI workloads. | medium | 100% |
| Ollama | product | Ollama released a preview offering faster performance on Apple Silicon devices, powered by Apple's MLX machine learning … | medium | 100% |
| NVIDIA | partnership | NVIDIA is accelerating Google's Gemma 4 family of open models for local agentic AI, enabling efficient deployment on edg… | medium | 90% |
| Anthropic | announcement | Anthropic's powerful AI model "Mythos" was revealed to the public through a data leak, exposing details about its capabi… | high | 100% |
| OpenAI | announcement | OpenAI is preparing to launch a new model named "Spud." | medium | 100% |
| Positronic Robotics | product | Positronic Robotics launched "PhAIL," an open-source benchmark to evaluate robotics foundation models on real hardware. | medium | 100% |
| Federal Trade Commission (FTC) | regulatory | A new bipartisan House bill proposes tasking the FTC with establishing transparency requirements for proprietary AI foun… | high | 100% |
| Meta | product | Meta's Llama 4 model faced mixed reception, and its flagship "Behemoth" model has been repeatedly postponed. | medium | 90% |
| Meta | product | Meta's unreleased AI model "Avocado" was delayed from March to May/June 2026 due to underperformance in internal tests. | medium | 100% |
| Anthropic | product | Anthropic accidentally leaked internal source code for its AI coding assistant, Claude Code, on March 31, 2026, due to h… | high | 100% |
| Anthropic | financial | Anthropic's Claude Code subscribers may face extra costs for OpenClaw usage, indicating potential pricing adjustments. | low | 80% |
| MLCommons | announcement | MLCommons released MLPerf Inference v6.0 benchmark results on April 1, 2026, introducing new LLM and text-to-video gener… | medium | 95% |
| OpenAI | product | OpenAI rolled out ChatGPT integration in Apple CarPlay, expanding its availability and use cases. | medium | 90% |
| OpenAI | product | OpenAI updated GPT-5 Pro to GPT-5.1 Pro on April 4, 2026, improving clarity and structure for users. | medium | 95% |
| Google Cloud | product | Google Cloud launched Veo 3.1 Lite, a new cost-effective video generation model for developers. | medium | 90% |
| Google Cloud | product | Google Cloud announced preview availability of Gemini 3.1 Pro in Vertex AI and Gemini Enterprise, enhancing reasoning ca… | high | 95% |
| OpenAI | product | OpenAI is iterating on GPT-5.4 variants, suggesting ongoing development and refinement of its flagship models. | low | 80% |
| xAI | announcement | xAI is targeting the release of Grok 5, indicating future model development plans. | medium | 85% |
| Anthropic | product | Anthropic's Claude Sonnet 4.6 achieved near-Opus level performance at Sonnet pricing in March 2026, offering high value. | medium | 90% |
| product | Google's Gemini 3.1 Pro model topped reasoning benchmarks in March 2026, demonstrating strong performance. | high | 90% | |
| Holo3 | product | Holo3's AI model reportedly surpassed OpenAI's GPT-5.4 in desktop automation tasks, demonstrating strong performance. | medium | 85% |
| PrismML | product | PrismML released a 1-bit AI model designed for iPhones, showcasing open-source innovation for edge devices. | medium | 85% |
| Alibaba | product | Alibaba's new Qwen3.6-Plus model topped OpenRouter's daily usage chart and ranked second globally in Code Arena for codi… | high | 95% |
| Anthropic | announcement | Anthropic confirmed the existence and ongoing testing of its unreleased Claude Mythos model, revealed via leak, with ear… | high | 95% |
| product | Google deprecated its Gemini 3 Pro Preview model on March 9, 2026. | low | 90% | |
| product | Google launched Nano Banana 2 and Gemini 3.1 Flash Image Preview on February 26, 2026, expanding its model offerings. | medium | 90% | |
| product | Google AI for Developers released `gemini-embedding-2-preview`, their first multimodal embedding model, on March 10, 202… | medium | 95% | |
| University of Cambridge | announcement | University of Cambridge introduced ADeLe, a new "ability profile" methodology to predict AI model performance on unseen … | medium | 90% |
| product | Google launched Gemma 4, a new family of open-weight AI models, on April 2, 2026, under an Apache 2.0 license. | high | 98% | |
| Arcee AI | product | Arcee AI released Trinity-Large-Thinking, a 399B parameter open-source model under Apache 2.0, offering customizability … | high | 95% |
| Stanford | announcement | A Stanford study revealed multimodal AI models like GPT-5 and Gemini 3 Pro can describe unseen images, exposing benchmar… | medium | 90% |
| product | Google released Gemini 3.1 Flash-Lite and Flash Live in March 2026, enhancing speed, real-time audio, and integrating "P… | high | 98% | |
| OpenAI | product | OpenAI released GPT-5.3 and GPT-5.4 models in March 2026, featuring expanded context, extreme reasoning, and a "Doer/Thi… | high | 98% |
WATCHER CONFIG
query: AI model release benchmark GPT Claude Gemini Llama open source foundation model
event_types: ['product', 'announcement', 'partnership']
impact_floor: all
time_window: 7
Signal event details
AI Model Releases
FIELD DATA
RAW DATA