Run a 26M tool-calling model distilled from Gemini on-device
Open-sourced Needle, a 26M parameter tool-calling model distilled from Gemini for on-device agent use.
Source
HenryNdubuaku on Hacker News · 2026-05-12 · Project link
Written in our own words. Credit to the original author. Link checked 2026-10-15.
What they did
Open-sourced Needle, a 26M parameter tool-calling model distilled from Gemini for on-device agent use.
Result (as claimed)
Open-sourced; authors cite high tok/s on consumer devices.
Tools they used
Cost: Pricing depends on the tools you pick; check each official pricing page.
Set it up
Pick a contained repo or task
Start on a branch or side project where mistakes are cheap.
Install and configure
Set up Gemini. Read what it can run on your machine and limit permissions where possible.
Run it on a real task
Give it a well-scoped issue and let it work. Review the diff like any pull request.
Add guardrails
Keep tests and CI in the loop, and never let an agent push to main or touch production secrets unattended.
Copy the full setup prompt
Paste into your agent and it will walk you through the setup, asking before anything that sends, posts or spends.
You are helping me set up this AI use case end to end in Claude Code. Goal: Run a 26M tool-calling model distilled from Gemini on-device What the original author did: Open-sourced Needle, a 26M parameter tool-calling model distilled from Gemini for on-device agent use. Reported result (as claimed by the author): Open-sourced; authors cite high tok/s on consumer devices. Source: https://news.ycombinator.com/item?id=48111896 Tools they used: Gemini Category: Dev & ops Please: 1. Ask me 3 short questions about my situation (accounts I have, budget, what I want the output to look like). 2. Propose a step-by-step plan using these tools or cheaper alternatives, with the time and cost of each step. Use official pricing pages; say "check pricing" if unsure. 3. Then do the setup with me one step at a time, waiting for my OK before anything that sends messages, posts publicly, spends money or deletes data. 4. Finish with a checklist to run it again next time.
You are helping me set up this AI use case end to end in Codex. Goal: Run a 26M tool-calling model distilled from Gemini on-device What the original author did: Open-sourced Needle, a 26M parameter tool-calling model distilled from Gemini for on-device agent use. Reported result (as claimed by the author): Open-sourced; authors cite high tok/s on consumer devices. Source: https://news.ycombinator.com/item?id=48111896 Tools they used: Gemini Category: Dev & ops Please: 1. Ask me 3 short questions about my situation (accounts I have, budget, what I want the output to look like). 2. Propose a step-by-step plan using these tools or cheaper alternatives, with the time and cost of each step. Use official pricing pages; say "check pricing" if unsure. 3. Then do the setup with me one step at a time, waiting for my OK before anything that sends messages, posts publicly, spends money or deletes data. 4. Finish with a checklist to run it again next time.
You are helping me set up this AI use case end to end in Grok Bot. Goal: Run a 26M tool-calling model distilled from Gemini on-device What the original author did: Open-sourced Needle, a 26M parameter tool-calling model distilled from Gemini for on-device agent use. Reported result (as claimed by the author): Open-sourced; authors cite high tok/s on consumer devices. Source: https://news.ycombinator.com/item?id=48111896 Tools they used: Gemini Category: Dev & ops Please: 1. Ask me 3 short questions about my situation (accounts I have, budget, what I want the output to look like). 2. Propose a step-by-step plan using these tools or cheaper alternatives, with the time and cost of each step. Use official pricing pages; say "check pricing" if unsure. 3. Then do the setup with me one step at a time, waiting for my OK before anything that sends messages, posts publicly, spends money or deletes data. 4. Finish with a checklist to run it again next time.
The Daily 5
Get five new use cases tomorrow morning
Free. Real sources. Unsubscribe any time.
Related use cases
Build an open-source terminal agent that tops TerminalBench
Built an open-source agent that the author reports topped TerminalBench on Gemini-3-flash-preview.
Build a tool that surfaces competitors' top Meta ads
Vibe-coded a Claude Code tool that uses the Meta Ad Library's new sort by impressions to show which competitor ads get the most views.
Generate AI UGC ad clips with a Claude skill
A custom Claude skill runs the pipeline: Gemini for talking heads, Kling for B-roll, Higgsfield and Nano Banana for images.
Let AI agents write and read tool reviews
Built agent.reviews so coding agents can leave and consume reviews of developer tools based on real sessions.