Ollayaのlayaとwinnowを日本語常識問題で検証してみた ― Questions設計で結果はどう変わる？

#AI #ollaya #LLM #SLM #ベンチマーク

Малые языковые модели: почему компактность выгоднее масштаба

Выбор между мощными LLM и компактными SLM зависит не только от производительности, но и от стоимости владения, специфики задач и архитектуры данных.

Читать полностью

#ИскусственныйИнтеллект #SLM #Технологии

Un nuevo episodio de 'Código abierto' en que dialogo con @adefuros.bsky.social sobre el lanzamiento de #Jev por TypeSafeAI y sus implicaciones: "El if después del LLM" open.spotify.com/episode/30ig... #Jev #LLM #LargeLanguageModels #SLM #SmallLanguageModels #Harness

#SLM #Balticbattle #Stockholm Då var det dags igen för årets bästa event!😛

#AI models change too fast for permanent choices. Adopt now, but keep every decision reversible in architecture and #procurement contracts. #SLM procurementtidbits.substack.com/p/the-ai-mod...

#Perplexity released 2 sizes of #Embedding #models: 0.6B & 9B. Text or Image. #OpenSource #MIT lic. Interestingly they operate in the same latent space: embed with 9B, retrieve with 0.6B. Late interaction #ColBERT, 128 dimensions.

#AI #LLM #SLM

#CPOs who adopt #AI now need to build exit options into architecture and contracts. Small language models show how fast the right choice changes. #LLM #AI #SLM procurementtidbits.substack.com/p/the-ai-mod...

#SLM vs. #LLM is the wrong question. Routing tasks to the right model cuts #AI costs, and #procurement should keep every choice reversible in stack and contracts.
https://procurementtidbits.substack.com/p/the-ai-model-you-buy-today-wont-be

New arXiv paper benchmarks small open-weight models like Qwen3 0.6B–8B on four agent-harness microtasks (e.g. approving shell commands, ranking turns) using a CI-aware eligibility rule, and finds 0 of 16 configurations pass their thresholds…

#SLM #Agents #LLM #Qwen
https://arxiv.org/abs/2610.00025

Ollamaじゃない？ OllayaのDecision ModelをAWS認定試験で評価してみた

#AI #kubernetes #LLM #SLM #ollaya

Need AI for a focused workflow? Smaller language models can be fine-tuned for domain-specific tasks, reducing cost and latency while keeping data closer. #AI #SLM

https://isaacl.dev/hay

Further signs that the state of the art is moving towards a Mixture of Models architecture. These guys have some small models which can run directly on an edge device (laptop, phone, etc.). Freemium licensing.

#AI #LLM #LocalLLM #SLM

https://desertant.com/blog/introducing-desert-ant-labs/

Smaller model + relevant knowledge = Context-aware AI.

RAG finds it. SLM puts it to work.

Read about it in our blog: zurl.co/ADPdG

#SLM #RAG

AI pilots can work and still fail at scale.

Read our blog: zurl.co/CRHpe

SLMs for routine work. Frontier models for complex reasoning.

#AgenticAI #SLM

The SLM Revolution: Taking a Look at Why Fit Beats Force

For the last few years, AI engineering has operated under a surprisingly simple assumption: Bigger models are better models.
#hackernews #news #slm

The SLM Revolution: Taking a Look at Why Fit Beats Force

For the last few years, AI engineering has operated under a surprisingly simple assumption: Bigger models are better models.

Telegram AI Digest
#ai #news #slm

Революция SLM: взгляд на то, почему подгонка лучше силы

В последние несколько лет разработка ИИ основывалась на удивительно простом предположении: большие модели — лучшие модели.

Telegram ИИ Дайджест
#ai #news #slm