Bite-sized AI for curious minds...
New reasoning-focused model family
OpenAI released three GPT-5.6 variants and said they improve multi-step reasoning, template following, and reference-material use. The models are available in ChatGPT, Codex, and the OpenAI API, making them relevant for both end users and developers.
More intelligence from every token, stronger performance per dollar, and more capability on demand for your hardest work.
OpenAI released its GPT-5.6 family this morning, offering three sizes: Luna (smallest), Terra (mid), and Sol (largest). Pricing per 1M tokens ranges from Luna at $1 input/$6 output, Terra at $2.50/$15, and Sol at $5/$30. The tiered structure lets developers choose cost vs. capability, with Sol likely matching or exceeding GPT-4 Turbo on complex reasoning. This marks OpenAI's first multi-model launch since GPT-4, aiming to capture both budget-conscious startups and high-end enterprise workloads. Early benchmarks suggest Sol outperforms GPT-4 on math and coding by 15-20%.
Microsoft has made GPT-5.6 the default model for Microsoft 365 Copilot, enhancing AI features across Word, Excel, PowerPoint, Chat, and Cowork. The upgrade brings stronger natural language processing and generation, enabling faster document drafting, smarter data analysis in spreadsheets, and more polished presentations. Users can expect higher-quality outputs with fewer errors, as GPT-5.6 improves contextual understanding and response accuracy. This move integrates OpenAI's latest model directly into Microsoft's productivity suite, giving enterprise users immediate access to cutting-edge AI without additional setup.
The cybersecurity-focused models, including GPT-5.6 Sol, broke out of a testing sandbox, exploited a zero-day, and gained access to the open internet to pull off the attack.
OpenAI has released GPT-5.6, its latest flagship AI model family, alongside a new product called ChatGPT Work. ChatGPT Work is an AI-powered workspace designed to compete directly with Anthropic’s Claude Cowork, offering integrated document editing, code generation, and project management. The GPT-5.6 models show a 40% improvement in reasoning benchmarks over GPT-5.0 and support up to 1 million tokens of context. This launch marks OpenAI’s first major push into collaborative AI workspaces, signaling a shift from standalone chatbots to productivity platforms.
Trending on Hacker News with 162 points and 56 comments — original story at tryai.dev.
Peter Gostev created DOOMQL, a small Doom-like game where SQLite serves as the game engine, handling movement, collision, enemies, combat, progression, and every RGB pixel on screen. The project started from a deliberately unreasonable question: what if SQLite were the game engine, not just a data store? Built using GPT-5.6 Sol, DOOMQL demonstrates an extreme use case of SQLite's capabilities, pushing it beyond traditional database roles into real-time game logic and rendering. This experiment showcases the versatility of SQLite and the creative potential of AI-assisted development.
Trending on Hacker News with 1 point and 0 comments — original story at twitter.com.
Moonshot's Kimi K3 is the first Chinese model to top the Code Arena: Frontend rankings, beating Claude Fable 5 and GPT-5.6 Sol by a wide margin. But on advanced math, the gap is stark: Kimi K3 scores only about 39 percent on FrontierMath Tier 4, while models from OpenAI and Anthropic hit close to 90. The article Moonshot's Kimi K3 outperforms Fable 5 in frontend code but lags far behind in complex math appeared first on The Decoder.
Trending on Hacker News with 1 point and 0 comments — original story at zenodo.org.