Skip to content
View shyringo's full-sized avatar
3 followers · 3 following

Achievements

Achievement: Starstruck

Achievements

Achievement: Starstruck

Block or report shyringo

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
shyringo/README.md

shyringo

Building local AI systems that bring large open models to laptop CPUs.
专注于让大模型在普通笔记本 CPU 上真正可用。

Native C · CPU inference · low-memory runtimes · reproducible performance

Open the Laptop LLM Finder

Featured projects

DeepSeek-V4-Flash-0731 in C

Run the native 284B-A13B DeepSeek-V4-Flash-0731 checkpoint on one laptop CPU. The pure C engine streams cold MoE experts from disk, has a tested 8 GB RAM path, needs no GPU or Python, and reaches 1.12 token/s in its best documented prompt-lookup workload. It supports terminal chat and a resident local OpenAI-compatible API with live token streaming.

DeepSeek-V4-Flash-0731 running through the native C engine on a laptop CPU

Qwen3.8-27B in C

Run Qwen3.8-27B locally on one laptop CPU with a native C engine, direct GGUF loading, an 8 GB RAM tested path and measured generation up to 2.52 token/s. Chat in the terminal or connect local apps through its resident OpenAI-compatible API. Runtime speedups preserve byte-identical full logits against the native baseline for the same GGUF.

Qwen3.8-27B running through the native C engine on a laptop CPU

Engineering focus

  • Native inference engines instead of wrappers around external runtimes
  • Low-bit CPU kernels, bounded memory planning and storage-aware execution
  • Reproducible TTFT, TPOT, memory and correctness evidence
  • User paths that go from clone to a real conversation without a GPU

Popular repositories Loading

  1. deepseek-v4-flash-0731-in-c deepseek-v4-flash-0731-in-c Public

    Run the native 284B-A13B DeepSeek-V4-Flash-0731 LLM locally on a single laptop CPU: pure C, 8 GB RAM minimum, no GPU, best TPOT 0.892 s/token, resident OpenAI-compatible local API. | 在笔记本单颗 CPU 上本地…

    C 16 1

  2. qwen3.8-27b-in-c qwen3.8-27b-in-c Public

    Qwen3.8-27B on one laptop CPU: up to 2.52 token/s, 8 GB tested, with no accuracy loss from runtime speedups. Resident OpenAI-compatible local API; native C, no GPU or Python. | 单颗笔记本 CPU 运行 Qwen3.8…

    C 3 1

  3. slime slime Public

    Forked from THUDM/slime

    slime is an LLM post-training framework for RL Scaling.

    Python

  4. Awesome-LLM-Inference Awesome-LLM-Inference Public

    Forked from xlite-dev/Awesome-LLM-Inference

    📚A curated list of Awesome LLM/VLM Inference Papers with Codes: Flash-Attention, Paged-Attention, WINT8/4, Parallelism, etc.🎉

    Python

  5. Awesome-LLM-Inference-Engine Awesome-LLM-Inference-Engine Public

    Forked from sihyeong/Awesome-LLM-Inference-Engine

  6. cpu-native-lm-train cpu-native-lm-train Public

    34.1M LM on a laptop CPU: 215,771 tok/s, 8,529x Qwen. QSELM beats Qwen3.5-0.8B on sealed long-document QA (90.6% vs 45.8%) and Qwen3-0.6B-FC on cross-turn agent memory (69.6% vs 3.2%). No GPU. | 笔记…

    Python