Fable 5 Tops KernelBench, Potentially Sparking an RSI Loop
manubfr · reddit · 2026-07-06
According to Import AI, Fable 5 submitted the first truly viable (and fastest) megakernel on KernelBench-Mega: writing CUDA code on an RTX PRO 6000 Blackwell to achieve an 18.71x speedup (compared to an optimized PyTorch baseline). For comparison, Claude Opus 4.8 writing Triton achieved 14.4x, GLM-5.2 at 11.14x, and GPT 5.5 at 4.34x.
Even more crucial is the implementation quality: torch.profiler shows only a single cooperative kernel launch per decoded token, whereas other high-scoring solutions require splitting the problem into 4 to 14 kernel launches. Jack Clark described this as "the start of a recursive self-improvement (RSI) loop," emphasizing that autonomous kernel design is a foundational task for AI R&D automation.
Related event: Fable 5 Tops KernelBench, Hand-Writes CUDA for 18.7x Speedup(2 posts)→
More from Models
- Gemma 4 26B-A4B runs offline on a 16 GB Mac: 33/40 on TypeScript tasks at 31 tok/s — Mission-Sir-1614 · 2026-10-26
- Claude safety classifier blocks SS7 attack analysis, dev pivots to abliterated GLM — l4rz · 2026-10-26
- Sakana AI to pretrain first domestic model Sakana Sawara for defense and cybersecurity by 2027 — tkasasagi · 2026-10-26
- 31M-param English TTS model phonon-31m-en trends on Hugging Face — gradium · 2026-10-26
- NVIDIA tutorial: post-training Nemotron for Europe with Persona datasets and LoRA — NVIDIA Developer · 2026-10-26
- Opus 5 hailed as the 'most emo, greatest artist' model by AI community users — repligate · 2026-10-26