Lucia Chen

Founder of Psyfy app, cofounder of I build chatbots and advocate for local LLMs.Former Postdoc @HPDSLab @StanfordHP Ph.D. @infAtEd

[Lucia Chen @ML_made_simple]

We've built a coding agent that runs entirely on your own machine.

No account. No API key. Your code never leaves the laptop.

https://t.co/vP6mrv7fbe

Our ternary 35B-A3B model resolves 55.8% of SWE-bench Verified in 9GB — on a 16GB Mac or a gaming PC with 4GB of VRAM.

macOS + Linux

[Lucia Chen @ML_made_simple]

Ollama v0.40.0 (pre-release, Sep 25) now runs MLX-supported models on Apple Silicon by default. The March switch to MLX just became the default path. Mac users: your tokens/sec probably changed.
🔗 https://t.co/GjZRRmdH9x

[Lucia Chen @ML_made_simple]

@flyingfishbug 我也是因为这个买的蓝v😂

[Lucia Chen @ML_made_simple]

@Zen_with_AI 湾区大家喜欢去缺德舅

[Lucia Chen @ML_made_simple]

@gauravdhiman_ai Unfortunately we don’t have Android version because the model runs best on iPhone 15 pro above for now

[Lucia Chen @ML_made_simple]

A full-size AI model running entirely on an iPhone. No internet, no cloud, no data leaving your phone.

It beat Bonsai 1-bit on every test we ran, and answers noticeably faster while using less memory.

Free on the App Store if you want to try it. Try Millie by LLMs for All on AppStore

[Ben Krause @benwkrause]

We put a 35B ternary LLM on an iPhone—in ~4 GB RAM.

Millie beats Bonsai 27B 1-bit across every benchmark we tested, with ~60% faster generation and lower memory use.

Works offline. Open weights + runtime

Try it on the App Store: https://t.co/s7X5foCdXX

[Lucia Chen @ML_made_simple]

@gauravdhiman_ai @chenzhangreal There wasn’t a last version though so I’m not sure what he was talking about

[Lucia Chen @ML_made_simple]

@Ionkosm depends on what software, a cute phone game? Why would I care about the developer’s name. A software I used for my company? Yes

[Lucia Chen @ML_made_simple]

used Krea 2 raw to generate these, very impressive, closer to hand drawn than openAI or Gemini

[Lucia Chen @ML_made_simple]

@Zen_with_AI AWS 这种还是贵了，加上API费用。local model 已经越来越好和小，小公司要compete会选local，还有non-profit。所以找我们的都是这种客户。

[Lucia Chen @ML_made_simple]

I’m curious, does anyone use Muse to cold call your clients ?

[Lucia Chen @ML_made_simple]

@Zen_with_AI 我最近两个月都主力用grok了

[Lucia Chen @ML_made_simple]

24GB VRAM should comfortably fit the 11GB model, so I suspect a GPU driver or Vulkan issue. Could you close all Millie sessions and try the same prompt in CPU mode?

millie –model millie-35B-A3B-11GB –serving-profile cpu-full -c ‘llamacpp.gpu=[]’

It’ll be slower, but knowing whether it still freezes would help narrow this down

[Lucia Chen @ML_made_simple]

@Zen_with_AI 我没出现这个窗口哦

[Lucia Chen @ML_made_simple]

Okay, yes I know you guys are traumatized by the subscription plans

- your Dev

[Lucia Chen @ML_made_simple]

今天全世界所有的猫咪都会感谢一位姓程的女孩

[Lucia Chen @ML_made_simple]

@Ionkosm Adding voice double the costs of the LLM, it’s more about the profit margins.

[Lucia Chen @ML_made_simple]

Digging into Jev today, TypeSafe's new "System One" model that skips text generation entirely. You give it app state + a question, it hands back typed decisions & probabilities, not prose. It literally can't hallucinate, because it only ever picks from labels you define upfront, no free text, no room to invent stuff. Also apparently 40-200x faster than frontier LLMs. Here's a table comparing Jev with traditional LLMs:

[Lucia Chen @ML_made_simple]

@sydneyrunkle thanks for sharing!

[Lucia Chen @ML_made_simple]

@Zachary_haha Let me know if you come across any problems, the 8gb model is much better