AI · alternatives · 2026

24 alternatives to Fast NF4 dequantization Triton kernel (1.41x faster than bitsandbytes)

"I wanted to see if I could optimize the dequantization bottleneck during 4-bit LLM inference. By writing a custom kernel in Triton to optimize memory access patterns, I managed to get up to a 1.41x speedup over the standard bitsandbytes implementation. Check out the source code and benchmarks, feedback is highly appreciated!"

Below are 24 products that do a similar job, ranked by how close each is in meaning and then by launch-day votes.

1
Speeding up LLM inference 2x times (possibly)▲419
Here's a project I've been working on for the last few months. It's a new (I think) algorithm, that allows to adjust smoothly - and in real time - how many calculations you'd like to do during inference of an LLM model. It seems that it's possible to do just 20-25% of weight multiplications instead of all of them, and still get good inference results. I implemented it to run on M1/M2/M3 GPU. The mmul approximation itself can be pushed to run 2x fast before the quality of output collapses. The inference speed is just a bit faster than Llama.cpp's, because the rest of implementation…
2024 · asciinema.org · its alternatives →
2
KVSplit – Run 2-3x longer contexts on Apple Silicon▲272
I discovered that in LLM inference, keys and values in the KV cache have very different quantization sensitivities. Keys need higher precision than values to maintain quality. I patched llama.cpp to enable different bit-widths for keys vs. values on Apple Silicon. The results are surprising: - K8V4 (8-bit keys, 4-bit values): 59% memory reduction with only 0.86% perplexity loss - K4V8 (4-bit keys, 8-bit values): 59% memory reduction but 6.06% perplexity loss - The configurations use the same number of bits, but K8V4 is 7× better for quality This means you can run LLMs with 2-3× longer…
2025 · github.com · its alternatives →
3
80% faster, 50% less memory, 0% loss of accuracy Llama finetuning▲385
Hi HN! I'm just sharing a project I've been working on during the LLM Efficiency Challenge - you can now finetune Llama with QLoRA 5x faster than Huggingface's original implementation on your own local GPU. Some highlights: 1. Manual autograd engine - hand derived backprop steps. 2. QLoRA / LoRA 80% faster, 50% less memory. 3. All kernels written in OpenAI's Triton language. 4. 0% loss in accuracy - no approximation methods - all exact. 5. No change of hardware necessary. Supports NVIDIA GPUs since 2018+. CUDA 7.5+. 6. Flash Attention support via Xformers. 7. Supports 4bit and 16bit…
2023 · github.com · its alternatives →
4
Tiny-vLLM – high performance LLM inference engine in C++ and CUDA▲205
Build your own high performance LLM inference engine in C++ and CUDA - a smaller version of vLLM - jmaczan/tiny-vllm
May 2026 · github.com · its alternatives →
5AT
A tiny LLM running at 21,000 tok/s on a $250 FPGA (Live Demo)▲79
A 3.16M-parameter INT4 transformer running entirely in the on-chip memory of a Xilinx Kria KV260. Zero DRAM in the token loop, 59,965 tok/s on the fabric, bit-exact. Chat with it live.
Aug 2026 · mikeayles.com · its alternatives →
6
We Beat MLPerf: Modern Storage for KV Offload and LLM Training▲35
High performance storage engine for efficient LLM inference and GPU Training.
Sep 2026 · theopenlake.com · its alternatives →
7
Running Gemma-4 26B at 124 tokens/SEC on a CPU, no GPU▲10
I wanted to know how fast a 26B mixture-of-experts model could run on a desktop CPU with no GPU. Got ~40 tok/s single-stream (lossless) and ~124 batched. The surprising part was the byte budget: for this model you compress the output head (32% of per-token bytes), not the experts (16%). The writeup has the bandwidth roofline and the dead-ends; the repo has the reproducible recipe. Happy to answer questions. Repo: https://github.com/arun-prasath2005/gemma4-cpu-moe
Jun 2026 · apeg.dev · its alternatives →
8
Outperforming VByte for Large Integers Using Phi-Encoding▲15
Properties of a Novel Binary Representation of Integers using Base-𝜙 - irradix/paper.md at main · DOSAYGO-STUDIO/irradix
2024 · github.com · its alternatives →
9
Run 500B+ Parameter LLMs Locally on a Mac Mini▲17
Hi HN, I built OpenGraviton, an open-source AI inference engine that pushes the limits of running extremely large LLMs on consumer hardware. By combining 1.58-bit ternary quantization, dynamic sparsity with Top-K pruning and MoE routing, and mmap-based layer streaming, OpenGraviton can run models far larger than your system RAM—even on a Mac Mini. Early benchmarks: TinyLlama-1.1B drops from ~2GB (FP16) to ~0.24GB with ternary quantization. At 140B scale, models that normally require ~280GB fit within ~35GB packed. Optimized for Apple Silicon with Metal + C++ tensor unpacking, plus…
Mar 2026 · github.com · its alternatives →
10LS
Llm.sql – Run a 640MB LLM on SQLite, with 210MB peak RSS and 7.4 tok/s▲8
Hi HN, I built llm.sql, an LLM inference framework that reimagines the LLM execution pipeline as a series of structured SQL queries atop SQLite. The motivation: Edge LLMs are getting better, but hardware remains a bottleneck, especially RAM (size and bandwidth). When available memory is less than the model size and KV cache, the OS incurs page faults and swaps pages using LRU-like strategies, resulting in throughput degradation that's hard to notice and even harder to debug. In fact, the memory access pattern during LLM inference is deterministic - we know exactly which weights are needed…
Apr 2026 · its alternatives →
11
Cuts Long Horizon Inference Costs by 50% via external KV Cache Offload▲22
Hey HN, we’re the developers of OpenLake, an open source storage engine for offloading LLM KV caches from GPU memory into a shared tier of RAM and NVMe. We built OpenLake because KV caches are outgrowing GPU memory. A single 256K token conversation on Gemma 4 31B produces approximately 43GB of KV state, more than half the memory of an 80GB H100. The problem becomes even harder across a cluster: a prefix cached on one GPU host is unavailable when the next request lands on a different GPU, forcing the new GPU to repeat work the fleet has already completed. Once the KV cache is offloaded,…
Jul 2026 · github.com · its alternatives →
12
Shoehorn, a library to quantize an LLM to fit your Mac's VRAM▲6
I made this after seeing someone posit the idea online yesterday over lunch then spent some time refining it. So far it's pretty impressive IMO! Right now I am running Qwen3-30B-A3B on my 24gb unified memory m4 MacBook Pro at 50 tok/sec and this should definitely not be working for such a large model on my middling hardware. Things are detailed in the README to get up and running and DESIGN.md has details on all the choices and such made along the way.
Aug 2026 · github.com · its alternatives →
13TF
The fastest way to run Mixtral 8x7B on Apple Silicon Macs▲18
I’d originally launched my app: Private LLM[1][2] on HN around 10 months ago, with a single RedPajama Chat 3B model. The app has come a long way since then. About a month ago, I added support for 4-bit OmniQuant quantized Mixtral 8x7B Instruct model, and it seems to outperform Q4 models at inference speed and Q8 models at text generation quality, while consuming only about 24GB of RAM[3] at 8k context length. The trick is: a) to use a better quantization algorithm and b) to use unquantized embeddings and the MoE gates (the overhead is quite small). Other notable features include many more…
2024 · its alternatives →
14LI
LLM Inference Calculator – Estimate VRAM, Latency, and Throughput▲6
LLM Inference Calculator — Estimate throughput, latency, TTFT, TPOT, and GPU memory usage for large language model inference. LLM 推理计算器 — 估算大模型推理吞吐量 (throughput)、延迟 (latency)、TTFT、TPOT 与 GPU 显存占用。
Aug 2026 · llm-inference-calculator-delta.vercel.app · its alternatives →
15
I shrank DeepSeek V4 Flash to 57GB and it wrote a compiler on my Mac▲21
I built a specialized package of DeepSeek V4 Flash 0731 (originally 284B total parameters, 13B active), preserving reasoning, tool calling and coding capabilities: https://huggingface.co/steadfastgaze/DeepSeek-V4-Flash-0731-... I let it write a minimal C compiler targeting ARM64, then test the result with Fibonacci and FizzBuzz programs, and it succeeded in less than 1 hour, with the full recording at: https://youtu.be/XiwSilmV8B0 You can run it on Silicon Macs with my engine https://github.com/steadfastgaze/MoEspresso, while one of the…
Aug 2026 · huggingface.co · its alternatives →
16
KTransformers–236B Model and 1M Context LLM Inference on Local Machines▲20
Hey Hacker News! We are excited to share our open-source project, KTransformers, a flexible framework designed for cutting-edge LLM inference optimizations! Leveraging state-of-the-art kernels from llamafile and marlin, KTransformers seamlessly enhances the performance of HuggingFace Transformers, making it possible to operate large 236B MoE models or extremely long 1M context locally with promising speed. KTransformers is a Python-centric framework designed with extensibility at its core. By implementing and injecting an optimized module with a single line of code, users gain access to a…
2024 · github.com · its alternatives →
17
Minimal LLM Post-Training Experiments on an 8GB GPU (SFT, DPO, GRPO)▲21
Minimal, readable LLM post-training experiments on one 8GB GPU. Measures forgetting, seed variance, and RL emergence. - pochenai/nano-llm-posttraining
Aug 2026 · github.com · its alternatives →
18
Beating GCC 12 – 118x Speedup for Jensen Shannon D. Via AVX-512FP16▲8
SIMD-accelerated distances, dot products, matrix ops, geospatial & geometric kernels for 16 numeric types — from 6-bit floats to 64-bit complex — across x86, Arm, RISC-V, and WASM, with bindings for Python, Rust, C, C++, Swift, JS, and Go 📐 - Release Beating GCC 12 - 118x Speedup for Jensen Shannon
2023 · github.com · its alternatives →
19WM
Which Mac runs which LLM, and how fast▲10
Pick a Mac and a model, get the memory it needs and the tokens/sec to expect. Covers M1 to M6, every unified memory size, Q4 to FP16. Formula fitted and validated on our own measurements on two machines.
6d ago · macyou.co · its alternatives →
20
50+ LLMs on 2 GPUs with 2-Second Swapping? We built AI-Native Runtime▲5
We've built InferX, a specialized runtime environment that fundamentally changes how LLMs are served. The core problem we solve is the latency bottleneck in AI inference, especially with large models. Current systems waste resources or suffer from painfully slow cold starts. InferX's AI-native architecture, with its "snapshot" technology, enables: * *Sub-2s cold starts:* Spin up models instantly. * *High density:* Serve more LLMs on the same GPUs. * *Optimal efficiency:* Maximize GPU utilization. This isn't just another API; it's a new execution layer designed from the ground up for the…
2025 · github.com · its alternatives →
21
Bonsai 1.7B ternary model at 442T/s on M4 Max▲13
We took a recently released Bonsai 1.7B ternary model from PrismML (https://github.com/PrismML-Eng/Bonsai-demo) and ran our agentic evolution search on it for 6 hours to optimize the Metal kernels. The search was fully autonomous. Measured against unmodified upstream llama.cpp at the same Bonsai/Q2_0 commit, same M4 Max: - tg128: 309.82 → 442.42 t/s (+42.0%) - pp512: 4250.32 → 4622.63 t/s (+8.8%)
May 2026 · agents2agents.ai · its alternatives →
22LR
Linear RNN/Reservoir hybrid generative model, one C file (no deps.)▲7
I just noticed it takes literally ~5 minutes to train millions parameters on slow CPU...but before you call Yudkowsky that "it's over", an important note: the main bottleneck is the corpus size, params are just 'cleverness' but given limited info it's powerless. Anyway, here is the project: https://github.com/bggb7781-collab/lrnnsmdds/tree/main couple of notes: 1. single C file, no dependencies. Below are literally all the "dependencies", not even custom header (copy paste from the top of the single c file): #define _POSIX_C_SOURCE 200809L #include #include…
Apr 2026 · raw.githubusercontent.com · its alternatives →
23
DualMix128 – A fast, simple PRNG passing PractRand (32TB) and BigCrush▲5
Hi HN, I'd like to share DualMix128, a pseudo-random number generator I developed. It's written in C and achieves very high speed while maintaining strong statistical properties for non-cryptographic use. GitHub (MIT License): https://github.com/the-othernet/DualMix128 Key points: * *Fast:* Benchmarked at ~0.36 ns per 64-bit generation on GCC 11.4 (-O3 -march=native). This was over 2x faster than `xoroshiro128++` (~0.74 ns) and competitive with `wyrand` (~0.36 ns) in the same tests. (Benchmark code/results in repo). * *Robust:* Passes the full TestU01 BigCrush suite…
2025 · github.com · its alternatives →
24
A benchmark + latency sim for LLM db queries: ClickHouse / Postgres▲12
Performance comparison of ClickHouse vs PostgreSQL using LLM-style query patterns on realistic aircraft tracking data - GitHub - 514-labs/LLM-query-test: Performance comparison of ClickHouse vs PostgreSQL using LLM-style query patterns on realistic aircraft tracking data
2025 · github.com · its alternatives →

Also compare

Ranked by how close each launch is in meaning, then by votes. Prices were read from each product’s own site when checked and can change. Refine with your own description →