Ciru.ai - Crown 👑

We can have our own entirely local AI assistants for research and work. This is CIRU AMDots: a voice interface for a general-purpose agent team, with the AI running locally on one AMD Ryzen AI Halo. AMDots is your own little AI crew, brought to life as colorful dots. Give the main one an idea and watch its workers spring into action, exploring, researching and creating together. Keep chatting while they work, then see what they’ve made, from useful discoveries to a narrated video. The brain is MiniCPM 4.7 Ultra , CIRU HALO AGENT, my custom Q8 conversion and optimized serving stack built on OpenBMB’s vision-language MoE: 35 billion total parameters / 3 billion active per token 262,144-token model context limit Custom Ciru vLLM + AMD ROCm+DSpark On the 395: ~1,969 prompt tokens/sec and 40.5 generated tokens/sec with a roughly 64K-token prompt 122.8 aggregate generated tokens/sec across eight concurrent vision/tool requests Those are separate model-serving benchmarks; the eight-request figure is combined throughput across requests. Model, runtime and benchmark details. The voice stack uses the rest of the hardware too: GPU: conversation and worker-model inference CPU: Moonshine Small Streaming English speech recognition, 123M parameters CPU + NPU: Pocket TTS, 100M parameters, with recurrent feed-forward operations offloaded to the XDNA 2 NPU Speech synthesis stays off the GPU, leaving that compute available for the agents. Voice and workers share the language model, with live conversation taking priority. Local model inference, a visual agent workspace, and research that can become something you can watch and hear. All on a small AMD desktop. AMD provided the Ryzen AI Halo system used for this project.

Codex is currently configuring GTA V in the background while I work. I told it to make the game look more realistic, but it kept pestering me for testing, so I finally told it to handle it itself! Now it's out there buying paid mods on its own (using its own agent card, through my bank).

I showcase Ming 3.0 Audio Flash, a 126B model running on my Ryzen AI Halo 395. Ming 3.0 Audio Flash is an audio understanding model. It can transcribe speech, identify sound effects and music, and describe the tone and emotion in people’s voices. You give it a recording and ask it questions about what it hears. In this demo I show: How it beats whisper v3 large in transcripts How you're able to ask it questions about what its hearing, emotion, music, sounds. There is a lot you can do with a model like this, and the fact that it's an open model is incredible. Thank you @AntLingAGI

Ling 3.1 Flash from @AntLingAGI has become very impressive in it's latest releases. Moving up the charts on the Artificial Analysis Intelligence Index. Particularly for Finance and Accounting. These models are very important when you need to get serious work done. I've been playing with other @AntLingAGI models in the last couple of days and really impressed. I plan to release more on that soon. Open Weight models for getting professional work done.

Ohhh, I'm looking forward to playing with this one! OpenBMB is one of the best if not the best at doing a lot with a little. A focused ocr jev model here could easily be best in class. Need to test.

Huge honor to be on this list. ❤️❤️❤️

I saw a few people test the new Nanbeige 33B but I only ever saw people talk about how slow it was. Of course it is slow, it's a dense model. Qwen 27b was also very slow at first before people spent time with spec decode and optimizations. I'm going to spend some time testing Nanbeige quality claims. If it's actually better than 3.8 27b then it could be worth more effort getting it working well. We will see.

Three more days until Qwen 4 LFG!

Local voice cloning is getting really good. I tested the newest Higgs model against @AntLingAGI new Ming 3.0 voice stack. Higgs has been my go to until now, really great quality and performance on local hardware. After this test, I am going to switch to Ming 3.0. Even the 500m model sounds better than Higgs in most of my tests. I'm getting a new mic tomorrow, sorry about the sound quality. I'll make some more videos later where I try out the entire stack, including the new Ming-3.0-audio-flash, a 126 billion parameter long-form audio reasoning agent.

What we all crave. Kimi K3.1 Padoga. Reminds me of Astra Ultra padoda. Not quite as epic though. This took 9% of my weekly $100 plan limit.

While other banks are banning agents Mercury Bank is giving my agents their own credit cards. 😎

A one shot of padoga's in padogas. A bit of chaos from 3.8 flash running locally on my AMD 395. ciru.ai/padoga-factory/garde…

Another from scratch model today! This is much more significant than another remake of Qwen. Plus, this one from an American lab @Recursive_SI Speed out of the box is promising, no optimization done yet except for tuning the config. Quality tests come next. For reference, this feels faster than initial Qwen3.6 35b when it first launched, which we later doubled in speed after optimizing for AMD hardware. If the quality is there, this is a big deal for local inference!

Today is a good day. A South Korean company (behind gaming hits like PUBG and Subnautica) has released a completely new, original model architecture and a model trained from scratch, building on existing innovations and leading-edge methods. It is incredibly expensive and challenging to train a competitive foundation model from scratch. Even giants like Google, with vastly greater resources, are being beaten on important benchmarks by this gaming company's first release. This is a much bigger deal than people realize. It looks competitive and perhaps even superior to Qwen's similarly sized MoE in several areas, even giving the much larger active-parameter dense Qwen3.8-27B some competition. They released it right away with NVFP4 and Apple Silicon MLX versions, under Apache 2.0. I think we're about to see some massive competition in open models, and I'm really excited about this. I'm going to begin making optimizations and builds for this model immediately.

It's a joke. I'm pissed. Give me $300 actual credit off my bill instead of $1000 of those fake usage credits any day.