The Open Weights
LatestModelsLeaderboardsCompanies
Subscribe
The Open Weights

The daily record of open-source AI. New model releases, leaderboards, and what's coming next — written for people who ship.

Refreshed every 12 hours

Discover

  • Latest releases
  • New today
  • Trending models

Browse

  • All models
  • Companies
  • Categories
  • Leaderboards

About

  • About
  • Editorial policy
  • Hardware estimates
  • RSS feed
  • llms.txt
  • Newsletter

© 2026 The Open Weights. An independent publication.

PrivacyTermsSMSAggregated by Claude · curated by humans.
LatestCactus Compute2
Cactus ComputeText / LLM

Needle2 Packs an Agentic LLM Into 14MB

Cactus Compute's tiny model brings tool and function calling to phones, wearables, and robots.

Jul 29, 2026
NotableApache 2.0
Needle2

Cactus Compute has released Needle2, a sub-1B-parameter language model that fits in roughly 14MB and is designed to run directly on constrained hardware like phones, wearables, and robots. Unlike most LLMs that assume a GPU or a cloud connection, Needle2 targets the edge, where memory and power budgets are tight.

What sets the model apart is its focus on agentic behavior. Needle2 supports tool and function calling, meaning it can be wired into external actions and APIs rather than just generating free text. That makes it a candidate for on-device assistants that need to trigger real operations without round-tripping to a server.

Why it matters

Running capable models locally addresses several persistent problems at once:

  • Latency: no network hop for each request
  • Privacy: data can stay on the device
  • Cost and reliability: no per-call inference bills or dependence on connectivity

A 14MB footprint is small enough to be embedded in firmware or shipped inside an app, which is where the wearable and robotics use cases become plausible. The model is released under the Apache 2.0 license, giving developers latitude to build and ship commercial products on top of it.

Needle2 is a text-and-code model, and Cactus Compute positions it as part of its broader push toward efficient on-device inference. More details are available at cactuscompute.com/needle. As always with tiny models, the practical question is how well the tool-calling holds up in real workflows — something developers will now be able to test directly.

Sources

  • Cactus-Compute/needle2

    Hugging Face

    Visit
  • Show HN: Needle2: 14MB agentic LLM for phones, wearables, smart home and robots

    Hacker News

    Visit
OlderLG AI Research debuts K-EXAONE 2.0, a 750B MoE modelLGAI EXAONE · Text / LLM · 4 months agoNewerNVIDIA's Nemotron VoiceChat 11B Targets Spoken AINVIDIA · Text → Speech · 4 months ago

Get the model

Hugging FaceHacker News

Specs

Context window2K tokens
ArchitectureNeedleForToolCalling
LicenseAPACHE-2.0
Downloads1.9K
Likes341

Modalities

Text / LLMCode

The Weekly Weights

Every open release that mattered, one email a week.

0 comments

No comments yet. Be the first to weigh in.

More from Cactus Compute

All Cactus Compute releases →
Mirage
Cactus Compute/Vision-Language

Cactus Compute's Mirage brings vision tasks on-device

The new open vision-language model targets detection, segmentation, and visual Q&A without relying on the cloud.

Nov 5, 2026
Whistle
Cactus Compute/Speech → Text

Cactus Compute's Whistle brings speech-to-text to the edge

A compact on-device ASR model built for edge hardware and WebAssembly, with early support for English, German, and French.

Sep 30, 2026
Cactus Needle 3
Cactus Compute/Text / LLM

Cactus Needle 3: tiny on-device tool-calling models

Cactus Compute's 8–29MB models aim to run automation and tool-calling entirely on-device, rivaling far larger cloud systems.

Sep 16, 2026

More in Text / LLM

All Text / LLM →
Aion-1.0-Instruct-2.3B
Microsoft/Text / LLMRuns anywhere

Microsoft's Aion 1.0 Instruct is a compact 2.3B chat model

A small, multilingual instruction-tuned model aimed at chat and summarization lands on Hugging Face.

Nov 18, 2026
SenseNova-Loop-9B
SenseTime/Vision-LanguageRuns on a laptop

SenseTime Releases Looped 9B Multimodal Model

The company's new SenseNova-Loop-9B pairs vision-language capabilities with a recurrent looped-transformer design.

Nov 17, 2026
Hunyuan Hy4-flash
Tencent/Vision-LanguageMulti-GPU server

Tencent releases Hunyuan Hy4-flash multimodal MoE

The lightweight, mixture-of-experts model adds vision and text instruction-following to Tencent's open Hunyuan lineup.

Nov 17, 2026