google/gemini-3.1-flash-lite-preview

91 views
Nick Antonaccio
Nick AntonaccioAdmin
Apr 13, 2026 at 14:59 (edited, 2 revisions)
#1

Together with pi and other little agentic tools, google/gemini-3.1-flash-lite-preview is a perfect fit. It is ridiculously fast - I haven't seen any faster model on an API. It's extremely knowledgeable for a flash model, capable of writing great code, and built specifically to run effectively in an agentic environment (so, it's not wordy and knows how to use tools). It's very cheap to use: $0.25/M input tokens $1.50/M output tokens - you have to really be fully involved burning tokens non-stop for many hours to spend a few dollars. So even for the lowest paying work, you can make money by using it to gets tasks completed faster, and save your own time to earn more and enjoy life more.

This is becoming more and more of a reality. What seemed incredible just a few months ago with Claude Code and the Claude models is now possible for a tiny fraction of the cost, no rate limits, and absolutely incredible speed.

It's hard to go back to a less performant model after you've experienced one that's so fast, smart and capable. I don't think anyone should be paying for Claude for all but the absolutely most demanding tasks that require every little bit of intelligence possible. The overwhelming majority of the time, that's just not the case. Keep in mind, /gemini-3.1-flash-lite-preview is better than just about anything that existed 1 year ago, and it's far stronger in most ways.

Don't get me wrong, I still use the zip file workflow in ChatGPT to build big software projects - I've never hit a rate limit doing that all day, days at a time, using it about as much a human could bear - and have never paid more than $20/month for all that extreme use - and by using that workflow, with the development ecosystems I favor (Python, Flask...), ChatGPT can accomplish virtually anything, nearly first shot, if development tasks are broken down into properly engineered iterative steps.

But, for all the agentic work - having an agent that can do things on your computer - that you can talk to, and give tasks to complete, to interact with other systems, gemini-3.1-flash-lite-preview is truly awesome.

Nick Antonaccio
Nick AntonaccioAdmin
May 08, 2026 at 17:17
#2

This model has been moved to production - the preview version still works, but there's now a version without the -preview ending.

Please login to post a reply.

© 2026 AI By Nick.