2026-09-16 04:28 UTC
Replicating Jev with Qwen-2.5
No summary available yet.
Topic
open-source
11 signals
4 in 24h
Open Source LLMs
Open weights, local models, model releases, and community inference stacks.
Latest Signals
Open Source LLMs feed
2026-09-16 00:55 UTC
Cerebras-powered, instant design canvas using Qwen3.8-27B [video]
Hacker News surfaced this AI signal from youtube.com: Cerebras-powered, instant design canvas using Qwen3.8-27B [video].
2026-09-15 23:37 UTC
Linux-Friendly Laptop Showcased for Local LLM AI Use
A video highlights a laptop with strong Linux support optimized for running local large language models (LLMs). The Hacker News discussion notes three key points but has no comments yet.
2026-09-15 21:17 UTC
Local LLM Engine Supports Tool Calls Despite Token Limits
A local LLM engine has been discussed that can handle tool calls even when truncated by token limits. The Hacker News discussion highlights this capability with two main points.
2026-09-15 01:37 UTC
Discussion on ROI for Local LLM Rigs on Hacker News
A Hacker News thread discusses the time it takes for a local large language model (LLM) rig to pay for itself, with 46 points and 96 comments. The conversation explores cost-effectiveness and investment recovery for running LLMs locally.
2026-09-14 21:14 UTC
Ollama 0.33.3 updates prompt_eval_duration metric
Ollama version 0.33.3 has changed the way the prompt_eval_duration metric is measured. This update was discussed on Hacker News with limited engagement.
2026-09-14 16:07 UTC
Show HN: Nari Qwen3-TTS and Qwen3-ASR – High accuracy, low latency and cost
Hacker News surfaced this AI signal from narilabs.com: Show HN: Nari Qwen3-TTS and Qwen3-ASR – High accuracy, low latency and cost.
2026-09-14 13:59 UTC
Notes on gotchas while migrating 35kb preprompts from Opus to self-hosted Ollama
Hacker News surfaced this AI signal from patrickmccanna.net: Notes on gotchas while migrating 35kb preprompts from Opus to self-hosted Ollama.
2026-09-14 01:04 UTC
DR-SL: A New Approach to Safe, Utility-Preserving De-identification for Cloud-Local LLMs
DR-SL (Dehydrate-Rehydrate with Self-Learning loop) is proposed to enable cloud-local LLM inference that keeps sensitive user data on-device while maintaining cloud-grade reasoning. It formalizes de-identification completeness to provide a release decision that balances safety and utility better th...
2026-09-12 09:15 UTC
Study Finds Local LLM Judges Consistent but Not Always Aligned with Human Ratings
Researchers evaluated local open-weight LLM judges LLaMA-3-8B and Qwen2.5-7B on 300 responses, finding that while these models produce consistent scores, they do not always agree with human evaluators. This highlights a gap between automated LLM evaluation and human judgment.
2026-09-02 21:00 UTC
GitHub Blog Explains New AI Terms Like Loops, Harnesses, and Squads
The GitHub Podcast breaks down emerging AI terminology such as loop engineering, harnesses, squads, and hill climbing that are appearing in developer discussions. These terms help clarify conversations around open source LLMs and AI development workflows.
Models (2)
View allAbout this Topic
Stable editorial lensOpen weights, local models, model releases, and community inference stacks.