#smalllanguagemodels — Public Fediverse posts

Kakao’s Kanana-2 Models Outscore Google, Alibaba on AI Safety

Kakao releases safety evaluation results for its lightweight language model, the Kanana-2 series. Photo courtesy of Kakao Kakao…
#EuropeSays #Korea #KR #Kakao #AISafety #AlibabaQwen #AssurAI #GoogleGemma #Kanana-2 #smalllanguagemodels
https://www.europesays.com/korea/123948/

Dispersion loss counteracts embedding condensation in small language models

https://chenliu-1996.github.io/projects/LM-Dispersion/

#HackerNews #DispersionLoss #EmbeddingCondensation #SmallLanguageModels #AIResearch #MachineLearning

Small Models and Smartphones Beat Sovereign AI GPU Clusters

ictworks.org/small-models-smar…

For most of the past three years, the implicit assumption in sovereign AI debates has been that meaningful AI capability requires hyperscaler-class compute. That assumption was correct in 2023. It is wrong now.

Two technical shifts have changed the geometry, and a third structural fact about the African device base decides whether the shifts can be put to work.

#AI #SmallLanguageModels #Africa #smartphones #SovereignTech

On June 9 2025 I wrote this and did nothing further with it aside from it becoming a draft toot, I *think*:

#SmallLanguageModels

Rather than search for the tag I'm posting it now and hoping clicking it turns up something of mine. Or someone else's.

SLMs are the future of Generative AI. LLMs will lose their moat due to open-weight systems, unified memory in Macs, and improved open source models. https://hackernoon.com/small-language-models-have-a-trillion-dollar-future #smalllanguagemodels

🚀 NEW on We ❤️ Open Source 🚀

Small language models are becoming a serious production choice, not just a compromise.

Nihal Kaul looks at how open source tooling helps teams fine-tune, serve, and deploy smaller models with more control.

https://allthingsopen.org/articles/small-language-models-open-source-ai-production

#WeLoveOpenSource #OpenSource #AI #SmallLanguageModels

Small language models do not just shrink the cloud. They end it. The sovereignty thesis becomes practical the day the model fits on the device.

https://mickai.co.uk/articles/small-language-models-finally-unlock-sovereign-ai

#smalllanguagemodels #slm #ondeviceai

SLM for home feed filtering.

ngl feels kinda necessary.

Not everytime there's a specific keyword I can easily filter for and block.

#sml #smalllanguagemodels #feelfiltering

New research shows that massive LLMs benefit from fine‑grained catalog context, delegating queries to specialized small models. This AI orchestration improves semantic understanding and intent routing, reshaping language model architecture. Dive into how the synergy between large and small models could boost efficiency and accuracy. #FineGrainedContext #AIOrchestration #IntentRouting #SmallLanguageModels

🔗 https://aidailypost.com/news/llms-need-fine-grained-catalog-context-large-model-routes-data-slms

A fine-tuned 3B model beat our 70B baseline. Here's why data quality and architecture innovations are ending the "bigger is better" era in AI. https://hackernoon.com/small-language-models-are-closing-the-gap-on-large-models #smalllanguagemodels

India’s SLM moment: Why Budget support is crucial for home-grown AI models

Budget expectations: Since its global rise just a few years ago, Generative AI has rapidly permeated every corner…
#NewsBeep #News #US #USA #UnitedStates #UnitedStatesOfAmerica #Economy #2026budget #Budget2026 #budgetandAI #budgetexpectations #BudgetNews #Business #India #indiabudget #smalllanguagemodels #UnionBudget2026 #unionbudgetofindia
https://www.newsbeep.com/us/416338/

In 2026, AI will move from hype to pragmatism

If 2025 was the year AI got a vibe check, 2026 will be the year the tech gets…
#NewsBeep #News #US #USA #UnitedStates #UnitedStatesOfAmerica #Artificialintelligence #AI #AIagents #ArtificialIntelligence #MCP #physicalai #smalllanguagemodels #Technology #worldmodels
https://www.newsbeep.com/us/385783/

In 2026, AI will move from hype to pragmatism

If 2025 was the year AI got a vibe check, 2026 will be the year the tech gets…
#NewsBeep #News #Artificialintelligence #AI #AIagents #ArtificialIntelligence #CA #Canada #MCP #physicalai #smalllanguagemodels #Technology #worldmodels
https://www.newsbeep.com/ca/384085/

Wondering where to start using small language models? Find top use cases where small language models would be better than large language models. https://hackernoon.com/when-to-use-small-language-models-over-large-language-models #smalllanguagemodels

What are small language models and how do they differ from large ones?
#Tech #AI #ArtificialIntelligence #LLM #SLM #LLMs #AIAssistants #SmallLanguageModels #TechNews #Microsoft #LargeLanguageModels #MachineLearning #The14 #The14Media
https://the-14.com/what-are-small-language-models-and-how-do-they-differ-from-large-ones/

https://winbuzzer.com/2025/11/25/microsoft-enables-local-computer-use-ai-agents-with-new-fara-7b-model-xcxwbn/

Microsoft Enables Local ‘Computer Use’ AI Agents With New Fara-7B Model

#AI #Microsoft #AIAgents #AgenticAI #Fara7B #OnDeviceAI #Automation #EdgeComputing #OpenSourceAI #MachineLearning #SmallLanguageModels

Learn how small language models are helping teams cut AI costs, run locally, and deliver fast, private, and scalable intelligence. https://hackernoon.com/cutting-ai-costs-without-losing-capability-the-rise-of-small-language-models #smalllanguagemodels

IBM's Granite 4.0 Nano Models Go Small, Promising Powerful AI on Your Laptop

#IBM #AI #Granite4Nano #OpenSource #OnDeviceAI #EdgeAI #SmallLanguageModels #MachineLearning #Efficiency #Innovation #BigTech

https://winbuzzer.com/2025/10/29/ibms-granite-4-0-nano-models-go-small-promising-powerful-ai-on-your-laptop-xcxwbn

Small language models are becoming increasingly practical in AI, offering efficiency and strong performance. This article explores seven leading models like Google's Gemma, Qwen and Phi-4, highlighting their strengths in areas such as reasoning, multilingual capabilities and accessibility. These models are reshaping AI by enabling on-device intelligence and versatile applications. #SmallLanguageModels #AI #MachineLearning #NLP #Gemma #Qwen #Phi4 https://www.kdnuggets.com/top-7-small-language-models

NVIDIA research challenges $57 billion AI infrastructure strategy with small language models: Small models demonstrate equivalent performance for 60-80% of enterprise AI tasks at fraction of operational cost. https://ppc.land/nvidia-research-challenges-57-billion-ai-infrastructure-strategy-with-small-language-models/ #NVIDIA #AI #ArtificialIntelligence #MachineLearning #SmallLanguageModels

Small Language Models Are the Future of Agentic AI

https://arxiv.org/abs/2506.02153

#HackerNews #SmallLanguageModels #FutureOfAI #AgenticAI #AIResearch #MachineLearning

It is about time we stop the use of big speak tables and instead try small speak tables.

I think the ten hundred words used for the Up Goer Five flying space car blue picture will work nice.

Up Goer Five blue picture:
https://xkcd.com/1133/

@theosanderson's nice Up Goer Five word look up thing:
https://splasho.com/upgoer5/

#LargeLanguageModels #LLM #SmallLanguageModels

There are two obstacles when you use #Assemblyscript and #Deno for #MachineLearning #ML and #SmallLanguageModels #SLM. 1. Assemblyscript works with linear memory and you have to flatten 3D #tensors and 2D #matrices into 1D #arrays for doing any matrix algebra. 2. Deno is a bit more complicated than Node.js when working with #WASM (#Webassemly) files. Here is how you do it. Webassembly makes NNUE models even faster in computation. This together with #WebGPU will become our future. #AI #javascript

Bigger isn’t always better, #SmallLanguageModels are gaining traction as cheaper, faster options that excel in targeted, domain-specific tasks. It’s about choosing the right-sized tool for the job. #AI #SLM #NLP #MachineLearning #ResearchSky

Small Language Models Are the ...

🤖 🗣️ Small language models are more reliable and secure than their large counterparts, primarily because they draw information from a circumscribed dataset. Expect to see more chatbots running on these slimmed-down alternatives in the coming months.

#SmallLanguageModels #Robots #Innovation

Read more: https://go.epfl.ch/b7s-en

Microsoft has expanded its Phi-4 AI lineup with Phi-4-mini and Phi-4-multimodal

#AI #Phi4 #MicrosoftAI #Microsoft #MachineLearning #SmallLanguageModels #OpenSourceAI #MultimodalAI #LLMs #GenAI

https://winbuzzer.com/2025/02/27/microsoft-expands-phi-4-small-ai-model-family-with-multimodal-and-mini-model-options-xcxwbn/

Key Points:
➡️ SLMs with 1-8B parameters can perform as well or better than LLMs.
➡️ SLMs are task-agnostic or task-specific.
➡️ SLMs balance performance, efficiency, scalability, and cost.
➡️ SLMs are effective in resource-constrained environments.
➡️ SLMs can be trained on consumer-grade GPUs.
➡️ SLMs include models like #Llama2, #Mistral, #Phi, and #Gemini.

https://arxiv.org/abs/2501.05465

#SLM #LLM #AI #MachineLearning #ArtificialIntelligence #Scalability #Performance #GPU #SmallLanguageModels

Hugging Face shows how test-time scaling helps small language models punch above their weight https://venturebeat.com/ai/hugging-face-shows-how-test-time-scaling-helps-small-language-models-punch-above-their-weight/ #AI #SmallLanguageModels

Hugging Face shows how test-ti...

[thread] Small language models
see also: https://en.wikipedia.org/wiki/Large_language_model
https://www.ibm.com/think/topics/small-language-models

* machine learning models
* processing, understanding, generating natural language content
* SLM more compact/efficient than LLM: large language models
* few million to few billion parameters vs LLM: 100B's - trillions
* parameters: internal variables that a model learns during training
* influence how model behaves/performs

#LLM #SLM #SmallLanguageModels #LanguageModels #NLP #ML #AI
#Microsoft #Phi3 #Phi4

Introducing Phi-4: Microsoft’s Newest Small Language Model Specializing in Complex Reasoning.

https://techcommunity.microsoft.com/blog/aiplatformblog/introducing-phi-4-microsoft%E2%80%99s-newest-small-language-model-specializing-in-comple/4357090
#ai #aimodels #phi4 #slm #smalllanguagemodels #cloud #azure #azureaifoundry

#EdgeComputing is like the cool kid on the tech block, handling data right where it’s created instead of sending everything back to the cloud.

The result? Faster decision-making, reduced bandwidth usage, and enhanced privacy.

But here’s the catch: edge devices often operate under strict constraints regarding processing power, memory, and energy consumption.

💡 Enter … #SmallLanguageModels (SLMs) - the efficient sidekick to save the day.

In this #InfoQ article, Suruchi Shah explores how SLMs can work their magic by learning and adapting to patterns in real-time, reducing the computational burden, and making edge devices smarter without asking for much in return: https://bit.ly/4fIQsDI

#AI #GenerativeAI #LLMs

Dives into the world of #GenerativeAI & #SmallLanguageModels - #SLMs!

"So latest trend in the language model evolution are the small language models or SLMs that offer many of the same benefits as LLMs, but they're smaller in size, they're trained using smaller data sets and they don't require a lot of computing resources." - Namee Oberst, Co-Founder of LLMWare

🎧 Listen now: https://bit.ly/4f9M5Sz

#AI #RAG #LLMs #EdgeComputing

While large language AI models continue to make headlines, small language models are where the action is. At least, that’s what Meta appears to be betting on, according to a team of its research scientists. #AI #SmallLanguageModels #SLM #Meta
https://tinyurl.com/mr3ja28s