Qwen3-0.6B megakernel: 527 tok/s decode on RTX 3090 (3.8x faster than PyTorch)
☆145Feb 10, 2026Updated 9 months ago
Alternatives and similar repositories for MegaQwen
Users that are interested in MegaQwen are comparing it to the libraries listed below. We may earn a commission when you buy through links labeled 'Ad' on this page.
Sorting:
- Aggressive decode optimizations for Qwen3-0.6B on RTX 5090☆64Feb 25, 2026Updated 8 months ago
- writing really fast kernels☆25Updated this week
- High-performance GPU kernels for LLM inference in OpenAI Triton. Fused RMSNorm, SwiGLU, INT8 GEMM with benchmarks and roofline analysis.☆45Jul 22, 2026Updated 3 months ago
- Tile primitives for speedy kernels☆21Sep 13, 2026Updated 2 months ago
- Kernels, of the mega variety :)☆883May 26, 2026Updated 5 months ago
- AI Agents on DigitalOcean Gradient AI Platform • AdBuild production-ready AI agents using customizable tools or access multiple LLMs through a single endpoint. Create custom knowledge bases or connect external data.
- ☆25Sep 23, 2026Updated last month
- A novel and efficient post-training INT6 quantization framework tailored for LLM inference.☆11Aug 29, 2026Updated 2 months ago
- Optimize GEMM with tensorcore step by step☆40Dec 17, 2023Updated 2 years ago
- An agent harness that compiles a model into one provably-correct, self-retargeting CUDA megakernel and self-tunes it past cuBLAS at batch…☆164Sep 18, 2026Updated 2 months ago
- mKernel: fast multi-node, multi-GPU fused kernels☆315Updated this week
- Writing FLUX in Triton☆42Sep 22, 2024Updated 2 years ago
- Pure Triton kernels for Qwen3.5-27B inference on NVIDIA B200☆125Feb 28, 2026Updated 8 months ago
- A PyTorch native automated mega-kernel compiler☆38Nov 12, 2026Updated last week
- A simple environment for writing and experimenting with hand-written CUDA PTX kernels.☆19Sep 11, 2025Updated last year
- Wordpress hosting with auto-scaling - Free Trial Offer • AdFully Managed hosting for WordPress and WooCommerce businesses that need reliable, auto-scalable performance. Cloudways SafeUpdates now available.
- Speed-of-Light SW efficiency by using ultra low-latency primitives for comms collectives☆13Apr 8, 2026Updated 7 months ago
- ☆98Mar 26, 2025Updated last year
- A PyTorch implementation of the GPT-OSS-20B architecture. All components are coded from scratch: RoPE with YaRN, RMSNorm, SwiGLU with cla…☆239Dec 2, 2025Updated 11 months ago
- ☆20Apr 26, 2026Updated 6 months ago
- ☆36Oct 15, 2026Updated last month
- Fastest kernels written from scratch☆630Aug 15, 2026Updated 3 months ago
- Region-level profiling for CUDA kernels with trace, NVBit, CUPTI, NSys, and an interactive Explorer.☆135Apr 17, 2026Updated 7 months ago
- CODA: Rewriting Transformer Blocks as GEMM-Epilogue Programs☆258Updated this week
- High-performance LLM operator library built on TileLang.☆204Updated this week
- Managed hosting for WordPress and PHP on Cloudways • AdManaged hosting for WordPress, Magento, Laravel, or PHP apps, on multiple cloud providers. Deploy in minutes on Cloudways by DigitalOcean.
- A series of high-performance GEMM (General Matrix Multiply) implementations Iteratively optimised for H100 GPUs in Pure CUDA.☆84Feb 18, 2026Updated 9 months ago
- Humming is a high-performance, lightweight, and highly flexible JIT (Just-In-Time) compiled GEMM kernel library specifically designed for…☆369Updated this week
- CUDA Craftax-Classic: 7x faster RL training than JAX☆80Sep 10, 2026Updated 2 months ago
- Implemented and benchmarked KV cache compression methods for LLM inference in Triton. Optimized kernels for KIVI and TurboQuant with Flas…☆18Jun 19, 2026Updated 4 months ago
- GGML bindings that aim to be idiomatic Rust rather than directly corresponding to the C/C++ interface☆20Sep 25, 2023Updated 3 years ago
- a guide to learn and implement inference techniques from scratch☆14Jun 18, 2026Updated 5 months ago
- ☆552Aug 26, 2026Updated 2 months ago
- TileFusion is an experimental C++ macro kernel template library that elevates the abstraction level in CUDA C for tile processing.☆119Oct 28, 2026Updated 3 weeks ago
- [NeurIPS 2025] ClusterFusion: Expanding Operator Fusion Scope for LLM Inference via Cluster-Level Collective Primitive☆79Aug 8, 2026Updated 3 months ago
- Proton VPN Special Offer - Get 70% off • AdSpecial partner offer. Trusted by over 100 million users worldwide. Tested, Approved and Recommended by Experts.
- ☆37Aug 31, 2026Updated 2 months ago
- A sparse-first inference engine (SparseEngine).☆110Nov 11, 2026Updated last week
- Optimize SGEMM step by step☆10Sep 4, 2024Updated 2 years ago
- 存放一些 CUDA 编程相关的博客文件。☆24Oct 16, 2025Updated last year
- Mirage Persistent Kernel: Compiling LLMs into a MegaKernel☆2,621Updated this week
- Decoding Attention is specially optimized for MHA, MQA, GQA and MLA using CUDA core for the decoding stage of LLM inference.☆48Jun 11, 2025Updated last year
- Triton kernels for Flux☆23Jul 7, 2025Updated last year
Are these results useful?