# @alxai_ — 2026-05-19

♥2 ↻0 · https://x.com/alxai_/status/2056788806431588651

Gemini 3.5 flash release - tldr ‘better’ and faster with ~300 tokens per second

Interesting to note the focus when discussing capabilities

Code & economically valuable work
(As imperfectly measured by evals Terminal Bench and GDP Val)

Still no clear way to get a sense of how ‘good’ a model way outside of imperfect benchmarks

![photo](../../../media/HIsug66a8AAo1Fl.jpg)
![photo](../../../media/HIsug68bMAARiJJ.jpg)
![photo](../../../media/HIsug68bsAAMkfr.jpg)
![photo](../../../media/HIsug69aMAAGjvw.jpg)
tags: author:alxai_, has-image, kind:image, kind:tweet, on:gemini-3-5-flash, year:2026
cited on: _dossiers/gemini-3-5-flash.md, gemini-3-5-flash