Meta's internal coding assistant Metamate is quietly using OpenAI's GPT-4 alongside its own Llama model, Fortune reports. The system dynamically pulls answers from either model based on the query. Even Mark Zuckerberg's philanthropic initiative is using OpenAI tech under the hood.
Andrew Ng’s team just released aisuite, an open-source Python library that lets developers switch between models from OpenAI, Anthropic, Ollama, and more by changing a single string. It wraps fragmented APIs into one standardized interface, drastically cutting the friction of building multi-model apps.
Elon Musk just announced that his startup xAI is launching an AI-powered game studio to "make games great again." He argued that big corporations have made current gaming too ideological, though he offered no concrete details on tech, games, or release dates.
Samsung just revealed Gauss2, its next-generation AI model family spanning three tiers: Compact for on-device computing, Balanced for speed and efficiency, and Supreme for heavy workloads. The models support up to 14 spoken languages and multiple coding languages, and already power internal tools like code.i. Samsung plans to weave Gauss2 into its wider consumer device lineup next.
Google just dropped Gemini-exp-1121, an experimental multimodal model that outperforms OpenAI's GPT-4o by roughly 20% across coding, math, and vision benchmarks. It features deeper contextual reasoning and optimized architectures, making it much sharper at turning raw design sketches into functioning code.
Vercel just upgraded v0 to build and run full-stack Next.js and React apps instead of just isolated UI snippets. It can now generate multiple files at once, pull in project environment variables, and deploy straight to Vercel. Basically, v0 is evolving from a UI generator into an end-to-end web dev sandbox.
Microsoft quietly assembled the enterprise agent market leader, revealing over 100,000 organizations now use Copilot Studio—doubling in just one quarter. At Ignite, it unlocked 1,800 third-party models from Azure beyond OpenAI, alongside new autonomous agents that handle complex workflows solo.
OpenAI just updated the ChatGPT desktop app for macOS with a new "Work with Apps" feature that automatically reads code from tools like VS Code, Xcode, and Terminal. Developers no longer need to copy and paste code back and forth to give the chatbot context. You will still have to paste its answers back into your editor yourself, though.
Anthropic just added a Claude-powered prompt improver and example management features to its developer console. The tool refines existing prompts using techniques like chain-of-thought reasoning and XML formatting in under a minute, lifting accuracy by 30% on a classification test.
IBM just upgraded its Quantum Heron processor and Qiskit stack, supporting up to 5,000 two-qubit gate operations and slashing a 112-hour benchmark run down to just 2.2 hours. The platform also integrates IBM Granite models to generate quantum code and optimize circuits with AI. Why it matters: Bringing generative AI into quantum supercomputing speeds up algorithm development for physics and materials science.
OpenAI is reportedly launching a new autonomous agent called Operator in January, according to Bloomberg. The tool can navigate web browsers to handle multi-step actions like booking flights or writing code on your behalf, rolling out as a research preview and developer API. It marks a clear pivot across big tech from conversational bots to software that actually gets work done.
Google just detailed how to deploy Gemma models on Dataflow using vLLM for streaming inference. By leveraging continuous batching and Dataflow's model manager, developers can spin up an end-to-end serving pipeline in minimal code, boosting inference throughput by over 23x in benchmarks.
The Qwen team just released the full Qwen2.5-Coder lineup, spanning six sizes from 0.5B to 32B parameters. The flagship 32B-Instruct model matches GPT-4o on major coding benchmarks like EvalPlus and Aider, supporting 40+ languages and 128K context. Most models run on an Apache 2.0 license, bringing serious frontier-level coding power directly to local hardware and Cursor setups.
OpenAI is rethinking its development playbook as its next flagship model, code-named Orion, shows smaller performance gains than anticipated. Early tests indicate it struggles to deliver dramatic improvements in areas like coding. The company is now leaning into synthetic data and post-training methods, while confirming Orion will not arrive this year.
Google just added native compatibility for Gemini models directly inside the OpenAI Library and REST API. Developers can now switch to Gemini using Chat Completions and Embeddings endpoints without rewriting code—simply point the base URL to Google and plug in a Gemini API key.
Google and Sourcegraph just shared test results from plugging Gemini 1.5 Flash's 1M-token context into the Cody coding assistant. The massive context dropped hallucination rates from 18.97% down to 10.48% on complex codebase tasks, while custom caching trimmed initial response latency from 40 seconds to roughly 5.
OpenAI just launched Predicted Outputs, a new API capability that slashes latency for gpt-4o and gpt-4o-mini by letting you feed in a reference string. It is tailored for partial rewrites—like updating docs, iterating on prior chat answers, or refactoring code inside an existing file.
NVIDIA just launched a new AI Blueprint for video search and summarization, enabling developers to build visual AI agents that analyze massive video archives and live streams using natural language prompts. Powered by vision-language models, the workflow cuts months off deployment for smart cities, warehouses, and traffic management.
Netflix's gaming division is betting big on generative AI to speed up development and unlock new experiences, calling it a "once in a generation" shift. The pivot comes just weeks after the company shut down its prestige gaming studio and laid off human developers.
Elon Musk's xAI just opened its API to the public, offering developers $25 in free monthly credits through the end of the year. The platform supports grok-beta with function calling and is compatible with OpenAI and Anthropic SDKs. Why it matters: At $5 per million input tokens, it is pricier than GPT-4o, but dropping in existing codebases requires almost zero friction.
