🐋 Run 284B DeepSeek-V4-Flash (<59GiB RAM, 44.4 tok/s pp, 9.1-14.7 tok/s) & 30B MoEs (25-111 tok/s) on a pure 64GB CPU at 256K ctx (0 GPUs). SwiGLU-H2, Stockfish-PVS L3 & 13ms Claude Code restore.
How fast synchopate/llama-modchip is gaining stars, and where it lands on Trendshift and GitHub Trending
Activity and trending
Stars, forks and merged pull requests each day or month, lined up with the repository's rank on Trendshift Trending
Oct 30 – Nov 8, 2026
Stars +319in 10 days
0100200300
290
Forks +19in 10 days
0510
7
Merged PRs 0in 10 days
01
Trendshift Trendingwhere the repository ranked each day it made the top 25
Not in the Trendshift top 25 in this period
Oct 30Nov 1Nov 3Nov 5Nov 7
The left and right arrow keys step from day to day; Home and End go to the first and the last
Stars, forks, merged pull requests and Trendshift Trending rank, by day
Day
Stars gained
Forks gained
Merged PRs
Fri, Oct 30, 2026
+0
+4
0
Sat, Oct 31, 2026
+7
+7
0
Sun, Nov 1, 2026
+2
+1
0
Mon, Nov 2, 2026
+3
+0
0
Tue, Nov 3, 2026
+1
+0
0
Wed, Nov 4, 2026
+0
+0
0
Thu, Nov 5, 2026
+1
+0
0
Fri, Nov 6, 2026
+1
+0
0
Sat, Nov 7, 2026
+14
+1
0
Sun, Nov 8, 2026
+290
+6
0
Total
+319
+19
0
1#1 on the list
7Rank that day
Not trending
Social mentions
Recent discussions about this repository across the web
5 posts
If results like this hold up, companies could run large models on machines they already own, with data that never leaves the building. Would your team run AI locally if the hardware bill dropped this…
Synchopate reports a 284.33B-parameter model running on pure CPU in 64 GB RAM, zero GPUs. The repo (llama-modchip v2.4.0) claims DeepSeek-V4-Flash, a MoE with 13.17B active params, fits at 56.56 GiB…
284b deepseek on a 64 gb cpu is on hn. did the math on the readme: 56.56 gib for 284.33b params = 1.71 bits per weight. the moe experts are iq1_s, ~1.56 bits. "lossless" there means same output as an…
Run the full 2.78-trillion-parameter Kimi K3 model, DeepSeek V4.1 Flash or GLM-5.3-Flash beyond available RAM by streaming activated weights directly from NVMe. A dependency-free, embeddable C inference engine.
Frontier-class LLM inference on a laptop CPU — gpt-oss:20b at ~110 tok/s on Apple M4 Max, 7.5x llama.cpp, no GPU. From-scratch NEON/SME kernels in Rust.
