Tech News
All News AI & ML Architecture DevOps Open Source Programming Team Management Testing & QA Web

Open Source

⚑ Report a Problem

Latest Open Source news from Tech News

All topics AI agents ai api architecture automation backend beginners career database devchallenge devops discuss javascript linux llm machinelearning mcp opensource performance productivity programming python react security showdev testing tutorial typescript webdev
All EN RU
EN

DeepSeek R1: The Open-Source Reasoning Revolution That Changes Everything

DeepSeek R1: The Open-Source Reasoning Revolution That Changes Everything The Reasoning Problem Traditional LLMs generate text token by token, left…

aideepseekmoereasoning
Dev.to Sep 13, 2026, 03:47 UTC
RU

CMF: новый формат для создания специализированных LLM моделей

CMF новый формат для нейросетей, чем CMF отличается от GGUF и SafeTensors, как несколько способностей можно накладывать на общую структуру без хране…

llmllm-моделиllm-архитектураllm памятьrustcargomoe
Habr Sep 4, 2026, 15:07 UTC
RU

[Перевод] Почему MoE-модель тормозит? Чек-лист из трех шагов

Вы запускаете большую MoE‑модель (например, Qwen, DeepSeek, Mixtral) на локальной машине через llama.cpp / llama-server , но скорость генерации токе…

moemixture of expertsartificial intelligenceискусственный интеллекттраблшутинг
Habr Aug 26, 2026, 13:55 UTC
EN

How a 176 KB C Binary Runs a 2.78-Trillion-Parameter Model on One CPU with 8 GB of RAM

The Problem Moonshot AI's Kimi K3 has 2.78 trillion parameters. Stored naively at bfloat16, that's 5,560 GB — more than the combined memory of two f…

aiopensourcesystemsmoe
Dev.to Aug 6, 2026, 01:21 UTC
RU

Kimi K3 — что реально даёт открытие весов самой большой модели

27 июля Moonshot AI выложила веса Kimi K3 — модели на 2,8 трлн параметров, которая до этого работала только через платный API. В…

kimi k3kimimoonshot aiоткрытые данныеmoeopen sourcellm
Habr Jul 28, 2026, 03:46 UTC
EN

2 TB of Ukrainian Law + DeepSeek V3 860B on GCP: What We'd Get

In production we have ~1.5 TB of full-text court decisions and their vector embeddings, plus another ~550 GB of other legal data: registries, legisl…

deepseekv3moetpuv5pgcp
Dev.to Jul 3, 2026, 21:35 UTC
EN

Step 3.7 Flash is a drop-in — except for one endpoint detail

Step 3.7 Flash shipped on May 29, 2026 as a structural upgrade to 3.5 Flash: same OpenAI-compatible SDK, new vision encoder, new runtime escalation,…

stepfunstep37flashllmmoe
Dev.to Jun 18, 2026, 09:36 UTC
EN

I built a Rust inference engine that streams MoE expert weights from NVMe SSDs, no GPU required

Most people trying to run Mixtral or DeepSeek-V3 locally hit the same wall: they don't have 80GB of VRAM. The common answer is "get better hardware.…

airustmoe
Dev.to May 27, 2026, 03:32 UTC

© Tech News — Headline Aggregator

English Русский
Sitemap Legal Notice Privacy Terms Copyright / Removal DSA Contact

Leaving the site

You are about to open an external website:

Continue →