AI tools · code autocomplete
Sweep 1.5B
Hacker Approved
A small open-weights model for predicting your next edit, praised for fast local responses and for beating larger models on this task. Critics say its suggestions are often overeager and sometimes delete needed code.
Summarized from 35 comments.
A small open-weights model for predicting your next edit, praised for fast local responses and for beating larger models on this task. Critics say its suggestions are often overeager and sometimes delete needed code.
- recommend
- 28
- criticize
- 3
- months recommended
- 6 of 13
Why people love it
Next-edit framing fits real refactors · 3
Nice work the next-edit framing matches how real refactors happen much better than token-level autocomplete.
Jan 2026 · commentThis seems to be a far more efficient method of accomplishing a more scoped problem.
Jan 2026 · commentNext edit makes a difference.
Jan 2026 · commentSmall model beats larger models · 3
training a relatively small model that can beat larger models for this important task
Jan 2026 · commentpushing the capabilities of _small_ models further and further
Jan 2026 · commentSweep 7B is 25% point better.
Jan 2026 · commentFast responses on local hardware · 2
I can even chat about C code with that model with LM Studio on my Macbook at 200 tokens per seconds
Jan 2026 · commentIt's really impressive so far, so quick to respond on a mac mini M2.
Jan 2026 · commentWorks well and saves time · 3
It's been working well so far (including next-edit prediction!)
Jan 2026 · commentI am super impressed by the performance/quality.
Jan 2026 · commentit really saves time and effort
Mar 2026 · comment
What people warn about
Suggestions are overeager or low quality · 2
I haven't been blown away by the quality of its suggestions unfortunately
Jan 2026 · commentIt's very overeager, suggesting tons of stuff that I never accepted because it made no sense.
Jan 2026 · comment
What hackers pick instead
- SupermavenFast, smart tab autocomplete that works well in editors like Neovim and PyCharm; critics say Cursor's purchase effectively killed it.10recommend
- Zeta 2 by Zed IndustriesFree open-weights autocomplete model from Zed that runs locally, so code stays on your machine. One critic finds it conservative, rarely proposing more than a few obvious lines.6recommend
Every quote
The 30 most detailed of 35 comments. Each links to the original.
Nice work the next-edit framing matches how real refactors happen much better than token-level autocomplete.
Recommends · Jan 2026 · in “Show HN: Sweep, Open-weights 1.5B model for next-edit autocomplete”
I can even chat about C code with that model with LM Studio on my Macbook at 200 tokens per seconds
Recommends · Jan 2026 · in “Show HN: Sweep, Open-weights 1.5B model for next-edit autocomplete”
training a relatively small model that can beat larger models for this important task
Recommends · Jan 2026 · in “Show HN: Sweep, Open-weights 1.5B model for next-edit autocomplete”
This seems to be a far more efficient method of accomplishing a more scoped problem.
Recommends · Jan 2026 · in “Show HN: Sweep, Open-weights 1.5B model for next-edit autocomplete”
I haven't been blown away by the quality of its suggestions unfortunately
Mixed · Jan 2026 · in “Show HN: Sweep, Open-weights 1.5B model for next-edit autocomplete”
It's really impressive so far, so quick to respond on a mac mini M2.
Recommends · Jan 2026 · in “Show HN: Sweep, Open-weights 1.5B model for next-edit autocomplete”
It's been working well so far (including next-edit prediction!)
Recommends · Jan 2026 · in “Show HN: Sweep, Open-weights 1.5B model for next-edit autocomplete”
pushing the capabilities of _small_ models further and further
Recommends · Jan 2026 · in “Show HN: Sweep, Open-weights 1.5B model for next-edit autocomplete”
Sweep 1.5B is "just" 12 % point better than Qwen2.5-Coder
Recommends · Jan 2026 · in “Show HN: Sweep, Open-weights 1.5B model for next-edit autocomplete”
I am super impressed by the performance/quality.
Recommends · Jan 2026 · in “Show HN: Sweep, Open-weights 1.5B model for next-edit autocomplete”
it really saves time and effort
Recommends · Mar 2026 · in “Emacs and Vim in the Age of AI”
Sweep 7B is 25% point better.
Recommends · Jan 2026 · in “Show HN: Sweep, Open-weights 1.5B model for next-edit autocomplete”
Next edit makes a difference.
Recommends · Jan 2026 · in “Show HN: Sweep, Open-weights 1.5B model for next-edit autocomplete”
It's very overeager, suggesting tons of stuff that I never accepted because it made no sense.
Criticizes · Jan 2026 · in “Show HN: Sweep, Open-weights 1.5B model for next-edit autocomplete”
AI is good at fixing a narrow subset of issues/tasks in GitHub, so humans still need to be involved in most tasks.
Mixed · Jul 2026 · in “Nothing works and everyone is euphoric”
It's an interesting model for sure but I am unable to get tab auto_completion/inline in zed
Mixed · Jan 2026 · in “Show HN: Sweep, Open-weights 1.5B model for next-edit autocomplete”
fascinating to see the transition from original concept to producing models. cool stuff.
Recommends · Jan 2026 · in “Show HN: Sweep, Open-weights 1.5B model for next-edit autocomplete”
We think SWE-Sweep is one of the most exciting research directions to work on right now.
Recommends · Oct 2026 · in “Show HN: Benchmark: AI doesn't find bugs unless you tell it what's wrong”
they have deleted code that I definitely still needed, and I had to reject those changes
Mixed · May 2026 · in “God Damn AI is making me dumb”
harness code that cleans up many bad behaviours that makes it a lot more usable
Recommends · Jan 2026 · in “Show HN: Sweep, Open-weights 1.5B model for next-edit autocomplete”
Sweep is designed to find these hidden leaks in your AI tool configurations.
Recommends · Apr 2026 · in “.env is safe but your –/.claude is not”
This is great and similar to what I was thinking of doing at some point.
Recommends · Jan 2026 · in “Show HN: Sweep, Open-weights 1.5B model for next-edit autocomplete”
This looks really neat, interesting technical writeup as well!
Recommends · Jan 2026 · in “Show HN: Sweep, Open-weights 1.5B model for next-edit autocomplete”
I've used it to write blog posts / large user-facing copy.
Recommends · Jan 2026 · in “Show HN: Sweep, Open-weights 1.5B model for next-edit autocomplete”
It's good. The blog post about it is very interesting.
Recommends · Jan 2026 · in “Show HN: Sweep, Open-weights 1.5B model for next-edit autocomplete”
so this plugin will be a great addon for me
Recommends · Jan 2026 · in “Show HN: Sweep, Open-weights 1.5B model for next-edit autocomplete”
This is a really good plugin.
Recommends · Jan 2026 · in “Show HN: Sweep, Open-weights 1.5B model for next-edit autocomplete”
This seems great for code
Recommends · Jan 2026 · in “Show HN: Sweep, Open-weights 1.5B model for next-edit autocomplete”
seems like a "why not/resume embellish/show VC" type release
Criticizes · Jan 2026 · in “Show HN: Sweep, Open-weights 1.5B model for next-edit autocomplete”
one has to sign in to their site to use the plugin
Criticizes · Jan 2026 · in “Show HN: Sweep, Open-weights 1.5B model for next-edit autocomplete”
Counts from 1,795,640 Hacker News comments, Nov 2025 to Nov 2026. How we count · Report a mistake