Stability AI just rolled out Stable Code Instruct 3B, an instruction-tuned coding LLM that beats CodeLlama 7B despite having less than half the parameters. It handles natural language queries, math, and code generation across popular languages. Why it matters: Developers can run powerful coding assistants on consumer-grade hardware, with weights ready to download on Hugging Face.
Google and Stack Overflow just announced a partnership powered by the new OverflowAPI. Google will integrate Stack Overflow's verified developer Q&A directly into Gemini for Google Cloud and its console, citing sources so coders can double-check answers. Meanwhile, Stack Overflow plans to build AI features on Google's Vertex AI.
GitHub just made Copilot Enterprise generally available for $39 per user per month. The new tier hooks directly into your company's private codebase and institutional knowledge, letting developers ask questions on GitHub.com, draft pull request summaries, and browse the web via integrated Bing search. It turns AI from a simple code autocomplete tool into an org-wide engineering partner, while promising never to train models on your private repos.
Apple is reportedly expanding internal testing of new generative AI features for Xcode, aiming to release a GitHub Copilot rival to third-party developers later this year. The company is also exploring chatbot-like search upgrades for Spotlight across iOS and macOS to handle complex app actions. After staying relatively quiet in the AI arms race, Apple looks ready to finally show its hand at WWDC.
Perplexity just partnered with Vercel to plug its online search models directly into the Vercel AI SDK. Developers building JavaScript and TypeScript web apps can now tap into pplx-7b-online and pplx-70b-online to fetch live web data without training cutoffs. Vercel also launched a multi-model AI playground alongside the update.
Abacus AI just released Smaug-72B, a fine-tuned version of Alibaba's Qwen-72B that is turning heads across the open-source community. It is the first open-source model to score an average above 80 on Hugging Face's Open LLM leaderboard, outpacing proprietary heavyweights like GPT-3.5 and Mistral Medium in key reasoning benchmarks. The weights are freely available now on Hugging Face.
Meta just rolled out Code Llama 70B, a larger coding model trained on 500 billion tokens with a massive 100,000-token context window. It arrives in base, Python-specialized, and instruction-tuned variants, all freely downloadable for commercial and research use under Meta’s permissive open license.
OpenAI just launched two new embedding models—text-embedding-3-small and large—bringing stronger retrieval benchmarks and adjustable dimensions to cut storage costs. It also slashed GPT-3.5 Turbo input pricing by 50% and shipped an updated GPT-4 Turbo preview to tackle code completion laziness. Why it matters: Developers get better RAG retrieval and cheaper tokens across the board.
GitHub just made Copilot Chat generally available for both Visual Studio Code and Visual Studio users. Powered by GPT-4, the conversational assistant lets developers explain code, spot vulnerabilities, and generate unit tests directly in the IDE—all included in existing Copilot plans at no extra charge.
Google just expanded its AI code assistance features to all Colab users, including those on the free tier in eligible regions. Users can generate Python code from plain English prompts and use an AI chatbot to explain and fix runtime errors. Why it matters: Bringing free AI pair programming to over 10 million active Colab users levels the playing field for students and researchers worldwide.
Google just launched the Gemini Pro and Gemini Pro Vision APIs for developers and enterprise customers. The models feature a 32K context window, function calling, and support across 38 languages. Devs can experiment for free via Google AI Studio with up to 60 requests per minute before paid tiers arrive next year.
Microsoft just released Phi-2, a 2.7B parameter language model trained on textbook-quality synthetic and web data. Despite its compact footprint, it matches or outperforms models up to 25x larger like Llama-2-70B on complex reasoning and coding benchmarks without RLHF. Why it matters: it demonstrates that strategic data curation can unlock emergent reasoning even in lightweight models.
A developer just rebuilt Google's heavily edited Gemini video demo using OpenAI's GPT-4V to prove real-time multimodal interaction is already possible. The custom setup correctly handled hand gestures, live drawings, and rock-paper-scissors in real time, with the open-source code now shared on GitHub.
AWS just launched Amazon Q, a generative AI assistant specifically tailored to enterprise data, systems, and AWS developer workflows. It handles everything from drafting code to diagnosing EC2 network errors while strictly respecting role-based permissions, and AWS won't use customer data to train underlying models. Why it matters: Cloud giants are finally addressing the enterprise security fears that have kept teams from adopting generative AI at work.
You.com just rolled out a set of APIs designed to give any large language model real-time access to the open web. Rather than short click-focused snippets from Bing or Google, it provides long-snippet indexes and RAG context starting at $100 per month. Anthropic, Cohere, and LlamaIndex have already integrated it.
GitHub just rolled out three AI-powered features for GitHub Advanced Security to help developers resolve vulnerabilities faster. Code scanning now uses LLMs to generate one-click fixes for JavaScript and TypeScript alerts right in pull requests, while secret scanning uses AI to spot unstructured leaked passwords that bypass standard rules.
Meta just partnered with Hugging Face and European cloud provider Scaleway to launch an AI startup program at STATION F in Paris. The initiative will mentor five early-stage startups building on open-source foundation models, giving them direct access to FAIR researchers, developer tools, and dedicated cloud compute.
Xbox just announced a multi-year partnership with Inworld AI to bring generative AI to game narrative and character design. The collaboration will produce an AI design copilot that turns developer prompts into scripts and dialogue trees, alongside an in-game runtime engine that dynamically generates quests and dialogue for players.
OpenAI just unveiled GPT-4 Turbo at DevDay, boosting context length to 128K tokens while slashing input prices by 3x and output by 2x. It also launched the Assistants API with built-in retrieval and code execution, effectively handling the messy state management developers previously had to build from scratch.
Phind just rolled out a new coding model based on CodeLlama-34B that scores 74.7% on HumanEval, outpacing GPT-4. Powered by Nvidia TensorRT-LLM on H100 GPUs, it delivers up to 100 tokens per second with a 16k context window. Why it matters: developers get GPT-4-level code generation at GPT-3.5 speed, completely open-sourced on Hugging Face.
