RT @vllm_project: 🎉 Herzlichen Glückwunsch an @JetBrains zu Mellum2-12B-A2.5B-Thinking, einem Open-Source-Modell mit 12B Parametern (Mixture of Experts), das nur 2,5B Parameter aktiviert und sowohl natürliche Sprache als auch Code mit einem Kontextfenster von 128K verarbeitet. Mellum2 läuft ab Tag 0 nativ in vLLM, mit Reasoning-Parser und Tool-Calling für agentic Workflows. 🔗 recipes.vllm.ai/JetBrains/Me… JetBrains (@jetbrains) Mellum begann mit Code-Vervollständigung. Mellum2 wurde für mehr gebaut – zur Verarbeitung von natürlicher Sprache und Code. Ein Open-Source-LLM mit 12B Parametern für Routing, RAG und Sub-Agents, optimiert für ultra-niedrige Latenz bei der Inferenz. Jetzt verfügbar auf @huggingface. Mehr erfahren: jb.gg/zpb9dp Video — https://nitter.net/jetbrains/status/2061444430884675791#m

mehr auf Arint.info

#AI #JetBrains #MachineLearning #Mellum2 #OpenSourceLLM #vLLM #arint_info

https://x.com/vllm_project/status/2061621691995005301#m

Poolside Laguna XS.2 and M.1: Agentic Coding Developer Guide 2026

Poolside released two new agentic coding models on April 28, 2026: Laguna XS.2, a 33B-total/3B-active open-weight MoE model under Apache 2.0 that runs on a single GPU, and Lagun...

https://wowhow.cloud/blogs/poolside-laguna-xs2-m1-agentic-coding-open-source-developer-guide-2026

#wowhow #poolside #agenticcoding #opensourcellm

This column is a good survey of where we're at right now and is written by someone working on actual open source models (ie. training data provided):

"The most successful open models will be complementary tools to closed agents. This is a path for open models to complement and accelerate the frontier of progress."

... "These models need to be almost brain-numbingly boring and specific. In a world dominated by coding agents, I want to build open models that Claude Code is desperate to use as a tool, letting its sub agents unlock entirely new areas of work. This is possible, but remarkably under-explored. Small models from the likes of Qwen and co. are still marketed on general-task benchmarks. The hype of “open models catching the frontier” distracts the world from this very large area of demand."

https://www.interconnects.ai/p/the-next-phase-of-open-models

#llm #openmodels #localmodels #opensourcellm

Alibaba's flagship Qwen team sees a key departure right after the open-source launch of Qwen 3.5. The move raises questions about the future of agentic inflection, AI workers and enterprise-grade LLMs. What does this mean for open-source AI? Dive into the details. #QwenAI #AlibabaAI #OpenSourceLLM #EnterpriseAI

🔗 https://aidailypost.com/news/alibaba-sees-key-qwen-ai-staff-exit-after-qwen35-open-source-release

Alibaba just released the Qwen‑3.5‑Medium model as open‑source, delivering Sonnet 4.5‑level performance on a single GPU. It uses a Mixture‑of‑Experts architecture and a new “Thinking Mode” to boost AI inference efficiency while staying lightweight. Dive into the details and see how this could reshape open‑source LLM development. #Qwen3_5 #OpenSourceLLM #MixtureOfExperts #ModelEfficiency

🔗 https://aidailypost.com/news/alibaba-open-sources-qwen35-medium-models-sonnet-45-performance

RE: https://hachyderm.io/@loleg/116041701680591995

Collected impressions of the #OpenSourceLLM summit on my personal blog https://log.alets.ch/121/