Jorick van Weelie
Marketing Lead at DataNorth | AI Enthusiast & Tech Storyteller
Articles written by Jorick van Weelie
Anthropic releases Claude Fable 5 and Mythos 5
Claude Fable 5 is Anthropic's first publicly available Mythos-class model, reaching 95.0 percent on SWE-bench Verified with a 1 million token context window and routing higher-risk requests to Claude Opus
OpenAI updates GPT-Rosalind
GPT-Rosalind is OpenAI's life-sciences model, now updated to use 31% fewer tokens than GPT-5.5 while improving accuracy on drug-discovery and laboratory research tasks.
Google releases Gemma 4 – 12B
Gemma 4 12B is an open-weight, encoder-free multimodal model with a 256,000-token context window that processes text, images, audio, and video on a 16GB laptop.
NVIDIA Releases Nemotron 3 Ultra and Cosmos 3
Nemotron 3 Ultra and Cosmos 3 together extend NVIDIA's AI model portfolio from inference hardware into the models themselves, offering developers open-weight alternatives for both language reasoning and multimodal Physical
Microsoft Launches MAI-Thinking-1 and MAI-Code-1-Flash
MAI-Thinking-1 and MAI-Code-1-Flash together signal that Microsoft is building its own model stack across reasoning, code, image, voice, and transcription, giving developers an alternative to OpenAI within the Microsoft ecosystem.
MiniMax launches M3
MiniMax M3 is an open-weight frontier model that combines 1M-token context, native multimodality, and competitive coding benchmarks at a price point well below Western proprietary alternatives.
NVIDIA Launches Cosmos 3
Cosmos 3 is the first fully open omnimodel that unifies physical AI reasoning, world simulation, and action generation in a single architecture.
Anthropic Releases Claude Opus 4.8
Anthropic launches Claude Opus 4.8 with improved coding benchmarks, dynamic workflows supporting 1,000 parallel subagents, and 3x cheaper fast mode.
Google Releases Gemini 3.5 Flash
Gemini 3.5 Flash is generally available as of May 19, 2026. Developers can access it through Google AI Studio, the Gemini API, Android Studio, Google Antigravity 2.0, Vertex AI, and
Google Veo: What is it and why is it important?
Google Veo is Google DeepMind's text-to-video AI with native audio and 4K output. The 2026 guide to features, pricing tiers, alternatives and business use cases.
Zyphra Releases ZAYA1-8B
ZAYA1-8B is an open-weight MoE reasoning model that delivers frontier-class math and coding performance with under one billion active parameters.
OpenAI releases GPT-5.5 Instant: 52.5% fewer hallucinations
OpenAI releases GPT-5.5 Instant as the new ChatGPT default with 52.5% fewer hallucinations, improved benchmarks, and enhanced personalization. Available to all users from May 5, 2026.
IBM Releases Granite 4.1: Open-Source Models
The release continues IBM's strategy of providing enterprise-grade open-source AI models with transparent training practices and permissive licensing.
Mistral AI Releases Medium 3.5
The release marks Mistral AI's largest dense open-weight model to date, combining frontier-level coding benchmarks with broad multimodal capabilities in a single model.
Alibaba Launches HappyHorse-1.0: the Number One Ranked AI Video Model
Alibaba's HappyHorse-1.0 is a 15-billion-parameter AI video model that generates synchronized audio and video in a single pass, currently ranked number one on the Artificial Analysis Video Arena.
OpenAI Releases GPT-5.5
GPT-5.5 is OpenAI's first fully retrained agentic model with a 1 million token context window, available now in ChatGPT and Codex for Plus, Pro, Business, and Enterprise users.
DeepSeek Releases V4: Open-Source 1.6T MoE Model with 1M Token
DeepSeek-V4 is a 1.6 trillion parameter open-source MoE model with a 1 million token context window that competes with frontier closed-source models at a fraction of the cost.
OpenAI Launches GPT-Image-2 with 99% Text Accuracy
GPT-Image-2 is a reasoning-native image generation model that sets a new benchmark for text accuracy, multi-image consistency, and complex scene composition in AI-generated visuals.
Moonshot AI Releases Kimi K2.6
Moonshot AI releases Kimi K2.6, a 1T parameter open-weight model scoring 58.6% on SWE-Bench Pro and 54.0 on HLE with tools.
Alibaba Releases Qwen3.6-Max-Preview
Qwen3.6-Max-Preview is Alibaba's most capable model yet, ranking first on six coding benchmarks while maintaining low inference costs through its 35B-total, 3B-active mixture-of-experts architecture.
