@QuixiAI: QuixiAI/ThunderKittens and QuixiAI/ThunderMittens are now rebranded to QuixiCore-CUDA and QuixiCore-Metal Announcing Qu…
Summary
QuixiAI rebrands ThunderKittens and ThunderMittens into QuixiCore-CUDA and QuixiCore-Metal, creating a unified family of cross-platform kernels for AI workloads.
View Cached Full Text
Cached at: 07/06/26, 01:59 AM
QuixiAI/ThunderKittens and QuixiAI/ThunderMittens are now rebranded to QuixiCore-CUDA and QuixiCore-Metal
Announcing QuixiCore - a family of kernels for every platform.
The idea is - all the repos will have the same set of kernels - developed and optimized for that platform. Enabling the same capabilities for all.
“Will mxfp8 work on my Intel Arc?” “Will BitNet work on my Gaudi?” “Will nvfp4 work on my mi300x?”
Yes, it will. 🆀
Similar Articles
@QuixiAI: https://x.com/QuixiAI/status/2073936537213915611
QuixiAI released QuixiCore, a family of native high-performance AI kernel libraries for modern accelerators, with standalone implementations for CUDA, Metal, ROCm, XPU, and Gaudi backends, all sharing a common contract but no shared code.
@QuixiAI: QuixiAI/ThunderMittens (fork from @HazyResearch) Porting ThunderKittens (and literally everything else) to Metal. Now w…
QuixiAI ported ThunderKittens to Metal, enabling kernel support on MPS and MLX for training models on Mac.
@QuixiAI: QuixiAI/ThunderMittens
QuixiAI announced ThunderMittens, a new AI model available on GitHub.
Qualcomm wants to grow in the AI space (multiple acquisition's underway)
Qualcomm is aggressively expanding in AI through multiple acquisitions, including Modular (creator of Mojo and MAX inference framework) and potentially Tenstorrent, signaling a significant push against Nvidia's CUDA ecosystem.
@hamzaelshafie: New in-depth blog post: "Dissecting ThunderKittens: Anatomy of a Compact DSL for High-Performance AI Kernels" This post…
A detailed blog post dissecting ThunderKittens, a compact DSL for high-performance AI kernels, including a bottom-up analysis of its abstractions and a benchmark implementing a non-causal attention prefill kernel that outperforms FlashAttention-2 by ~1.55x and matches FlashAttention-3.