@charles_irl: Added a fun lil widget to the LLM Engineer's Almanac -- a "Token Timing Simulator" so you can get a visceral feel for w…
Summary
A token timing simulator widget was added to the LLM Engineer's Almanac, demonstrating the DFlash technique achieving ~1k TPS, to help users viscerally understand benchmark performance numbers.
View Cached Full Text
Cached at: 05/15/26, 07:07 PM
Added a fun lil widget to the LLM Engineer’s Almanac – a “Token Timing Simulator” so you can get a visceral feel for what a benchmark perf number means. Here’s @dcw02’s latest work with @zhijianliu’s DFlash technique in @sgl_project – ~1k TPS! https://modal.com/llm-almanac/token-timing-simulator…
LLM Engineer’s Almanac - Token Timing Simulator
Source: https://modal.com/llm-almanac/token-timing-simulator TextInput tokensOutput tokens
Similar Articles
Getting a feel for how fast X tokens/second really is.
The author introduces a web-based script designed to help users intuitively understand token-per-second speeds in local LLM setups by simulating text, code, and reasoning generation rates.
@charles_irl: This block quant visualizer is another page in our LLM Engineer's Almanac -- a one-stop shop for engineers looking to o…
A new page in the LLM Engineer's Almanac provides a block quant visualizer to help engineers understand quantization formats for owning their LLM inference.
How fast is N tokens per second really?
A web tool that lets users visually experience different LLM token generation rates (e.g., 5–800 tok/s) across code, text, reasoning, and agent modes, helping internalize performance numbers from benchmarks.
@LottoLabs: This is awesome work Dflash for qwen 3.5/6 series
Charles Frye announces the co-release with Z Lab of six new DFlash speculators for Alibaba Qwen 3.x models, achieving over 1k output tokens per second for Qwen 3.5 122B-A10B on a B200.
Qwen-3.6-27B, llamacpp, speculative decoding - appreciation post
Reddit user demonstrates llamacpp speculative decoding boosting Qwen-3.6-27B token speed from 13.6 to 136.75 t/s, sharing exact commands and hardware setup.