I ran Gemma 4 (26B) on a 10-year-old-GPU, and it's reliable enough to replace the cloud
…So, I ran the So, I ran the model with ./llama-server -m "/root/models/gemma-4-26B-A4B-it-Q4_K_M.gguf" -c 65536 -ngl 999 --n-cpu-moe 40…
Search
Related topics:
Claude
Top stories
I ran Claude Code and Cursor together, and the combination does something neither can do alone
1d agoClaude Design looks impressive until you actually try to use it
1d agoClaude Code out of the box is good, but these mods make it actually production-ready
4d agoI replaced ChatGPT, Claude, and Gemini on my phone with a local LLM, and it's a mobile upgrade I didn't expect
1w agoI switched from VS Code's marketplace to Open VSX, and Microsoft's is the one falling behind
…The registry provides standard marketplace features, including extension discovery, installation, versioning, and updates. You can browse available extensions, install them directly within supported editors, and receive updates through the same extension management…
I connected my local LLM to Home Assistant through MCP, and now my smart home manages itself
…My LM Studio models now control my smart home I initially tried to control Home Assistant using some low-parameters models, but they left a lot to be desired. The DeepSeek R1…
ChatGPT can now remember you better by dreaming about you while it 'sleeps'
…For a while, AI developers advertised their models as a reliable worker that's always on, but recently, we've seen some companies announce that, actually, it's best if AI models…
Home Assistant's local LLM support outperforms Gemini for Home, and Google knows it
…You’re at the mercy of the model Google chooses and wait for an update if it falls short. Running LLMs locally frees you from those constraints — of course, you still need…
I talk to Perplexity Computer all day, and it's finally what AI assistants promised to be
…Perplexity's Computer brings the best LLMs to one platform Over 20 models, one tool Agentic AI tools, where LLMs can use tools on your behalf rather than simply giving you instructions…
Tines built the better AI workflow builder, but n8n is the one self-hosters will actually use
…candid about why generation runs on Anthropic models only. The prompts use Anthropic's cache_control for cost efficiency, and the add_nodes and update_parameters tools rely on passthrough() schemas that…
XDA's first dedicated newsletters are here: Meet AI Insider and Maker Weekly
…Staying true to our roots at XDA, we'll also have technical breakdowns when major models or updates drop. There are thousands of AI newsletters out there. A good chunk of them…
I started using my local LLMs and an MCP server to manage my NAS – it's surprisingly powerful (and safe)
…Sign in to your XDA account Despite my general distaste for shoehorned AI features that nobody wants, I must admit that large language models have boosted my productivity quite a bit. And…
I moved my entire research workflow to Google AI Pro, and I'd do three things differently if I started over
…It's just unintuitive if you're coming from ChatGPT or Claude expecting the same mental model.