The bigger file is the faster one. 2.16x faster at prompt processing, with 21 percent…
This is a AI post classified by Jev as AI infra & evals (a model release), kept by the AI Radar because it carries real work, not commentary.
The bigger file is the faster one. 2.16x faster at prompt processing, with 21 percent more bytes in it. Bonsai 2 27B went up today in two packings. PTQ1_0 packs trits dense, 1.75 bits per weight, 5.95 GB. PQ2_0 gives every trit its own 2 bit slot, 2.13 bits, 7.21 GB. Same weights, same machine, RTX 5090, from their own table. Prefill 3893 tok/s against 1805, decode 129.9 against 120.5. Decode is not one story across that whole table. A100 goes 73.9 down to 54.7 when the file gets 17 percent smaller, so 26 percent slower for less traffic. Four of eight cards gain, four lose. Unpacking dense
Posted by 코지베어 🐻 CozyBear (1.2k followers) 1 days ago · 0 likes · 14 views · view the original post on X. Kept by the AI Radar as AI infra & evals.
More AI work like this
- Trains your own large language model from scratch using plain PyTorch. — @tom_doerr
- Here's a useful use-case for @typesafeai for a change. — @iam_zachi
- This is genius. — @daniel_mac8
- Which GPU should you use for embedding workloads? — @runpod
- Big update to the @Alibaba_Qwen Qwen3.8-Flash-Next single DGX Spark recipe! — @jvr0x
- 🔰 The Databricks Advanced Learning Festival runs September 16 through October 14, with… — @databricks
- Q: What is a trace? — @HamelHusain
- Inference scaling part 1. — @rasbt
Every post is read and classified by Jev (TypeSafe): what it is, which market it belongs to, and whether the link is a real tool. 27.2k posts from 4.8k X accounts over the last 14 days, 1.2k tools, 19 markets. Collected every 5 minutes, fully re-ranked every hour — last update 2026-09-19 18:50 UTC. Full method.