🧠 OpenAI finally launches open-source models 🧠
First Apache 2.0 licensed models since GPT-2 in 2019

OpenAI unveiled gpt-oss-120b and gpt-oss-20b, its long-awaited open-weight reasoning LLMs that match or exceed o4-mini and o3-mini in performance.

The 120B variant performs on par with o4-mini on core benchmarks while being deployable on an 80GB GPU, while the smaller 20B version runs on laptops with 16GB of memory. Both models feature adjustable reasoning levels (high, medium, low) and support agentic workflows with function calling, web search, and Python execution.

This marks OpenAI's first open models since GPT-2, instantly becoming #1 among 2M models on Hugging Face.

🤖 Anthropic releases Claude Opus 4.1 🤖
Incremental upgrade boosts coding and research performance

Anthropic released Claude Opus 4.1, an incremental upgrade improving performance across real-world coding, in-depth research, and data analysis tasks. The model shows notable coding improvements, boosting performance on SWE-bench Verified from 72.5% to 74.5% and excelling at multi-file code refactoring and identifying correlations in codebases.

Customers report real-world gains in tasks requiring attention to detail and agentic actions. Available to paid users and businesses, Anthropic says this upgrade marks the start of "substantially larger improvements" planned for its models.

🌍 Google's Genie 3 interactive world model 🌍
Real-time environment generation with 24fps consistency

Google DeepMind announced Genie 3, a general-purpose world model that generates interactive environments in real-time from text prompts with surrounding and character consistency. Users can generate unique 720p environments with real-world physics and explore them at 24fps, with the model's visual memory extending up to one minute.

Genie 3 enables users to change worlds dynamically by inserting new characters, objects, or altering environment dynamics entirely. This breakthrough lays the foundation for scalable training of embodied AI, where machines can tackle "what if" scenarios by adapting in real time.

INTERESTING TO KNOW

🎵 ElevenLabs introduces Eleven Music generation 🎵

ElevenLabs introduced Eleven Music, its multilingual music generation model with control over genre, style, and structure, plus the option to edit both sounds and lyrics. The model represents ElevenLabs' expansion beyond voice synthesis into comprehensive audio generation, offering creators precise control over musical output. Users can generate music in multiple languages while maintaining consistent quality across different genres and styles. This launch positions ElevenLabs to compete directly with specialized music AI platforms while leveraging their expertise in audio generation.

🛒 Shopify adds agent-focused commerce features 🛒

Shopify added new agent-focused features including a checkout kit to embed commerce widgets into agents, low-latency global product search, and a universal cart system. These tools enable AI agents to seamlessly integrate commerce functionality, allowing users to shop directly through conversational interfaces. The checkout kit provides developers with pre-built components for rapid commerce integration, while the global product search ensures fast, accurate results across Shopify's entire ecosystem. This move positions Shopify as the infrastructure backbone for the emerging agentic commerce landscape.