Red Dead Redemption 2 is running at 30+ FPS on an iPhone 18 Pro. No PC, no streaming, fully on-device via Madeira. x86-64 → arm64 with FEX Windows → iOS syscalls with Wine DX12 → Metal with the madeira-d3d12 runtime What AAA game should we get running next?

🚀Excited to announce Nanbeige4.3-33B! To push the performance ceiling of small agents, our strategy in Nanbeige4.3 starts with building a powerful teacher first. Nanbeige4.3-33B delivers SOTA agentic performance among open models at ~30B parameters: * Strong results on agentic coding and cowork benchmarks. * When integrated into mainstream agent harnesses, it tackles real-world tasks with cost-efficient inference. * It will serve as a teacher for upcoming smaller Nanbeige models, starting with Nanbeige4.3-3B. huggingface.co/Nanbeige/Nanb… huggingface.co/Nanbeige/Nanb… #LLM #Agent

An iPad just ran a PS5 binary. Not streaming. Not Remote Play. No PS5 in the room. No console emulator. The PS5 build of Dreaming Sarah, relinked by AnyPS5 and booted on an M2 iPad: walk, jump, talk to NPCs, music, touch controls. Here's how it works 🧵

Red Dead Redemption 2 is now running on an iPhone 18 Pro. No PC, no streaming, fully on-device. Translating DirectX 12 to Metal with custom runtime that leverages DXMT's airconv for SM5 shader conversion. This is an experimental build of Madeira with graphical glitches.

Play Stray on your iPhone with Madeira. 🐈‍⬛ The game runs directly on-device. No cloud gaming. No remote PC. No streaming. Windows PC games running on iOS.

Dead Space just ran on an iPhone. Not through cloud gaming. Not through Remote Play. Fully on-device. Madeira brings Windows PC games to non-jailbroken iPhones. It combines: • Wine WoW64 • FEX x86-64 translation • DirectX 9 → Metal (DXMT) • JIT via StikDebug

The rumours are right. 0xalpha is absurdly good. It figured out an issue Fable and Sol have been kicking back and forth for a few hours now.

Performance of Qwen3.8-27B:

As yet unverified DeepSeek V4 Pro benchmarks from WeChat. Seismic if accurate.

Whalesign!

Unless we discover problems after mission data review, SpaceX will attempt to catch the ship with the tower on next flight

Today we're releasing Laguna S 2.1, our most capable model to date. It's a 118B total parameter Mixture-of-Experts model with 8B activated per token, a context window of up to 1M tokens, and thinking and no-thinking modes. Capable enough to hold its own against models many times its size. Small enough to run on a single @NVIDIAAI DGX Spark. Laguna S 2.1 is fully open under OpenMDW-1.1, with weights available today on @huggingface poolside.ai/blog/introducing…

Announcing our $130M Series A to build the Open Superintelligence Stack Led by Radical Ventures, with NVIDIA, Intel Capital, Dell Capital, and existing investors Train, deploy, and continuously improve your own models using our stack. Own your intelligence.