Wild Edge monitors every inference, captures what degrades, and builds a dataset from real failures and user feedback. Automatically.
Warsaw
Joined April 2026
- Tweets63
- Following153
- Followers13
- Likes106
WildEdge retweeted
Today @WSJ covered our work with @MercedesBenz. our flagship LFMs are now rolling out in vehicles in North America, running directly on the hardware inside the car to power in-car assistance! 🌊
this is one of the first deployments of local generative AI in the real world at scale! ✨
seeing our technology become part of someone’s everyday life is one of the most rewarding feelings of building Liquid.
years of work become a small, useful part of someone’s day. they don’t need to know which model is running. they just need it to understand them and work seamlessly.
this is really hard and it took us years of fundamental work from first principles to made happen with a brilliant team of engineers from Mercedes and Liquid.
I believe on-device AI will fundamentally change our relationship with the technology around us. embedded AI is about change of ownership.
With Mercedes-Benz, we are now in production and are building toward deployment at scale. The on-device AI revolution is taking shape in products people actually use, and I’m proud that @liquidai is helping lead it.
To our team and our partners at @MercedesBenzUSA and HQ: thank you for the care, persistence, and trust it takes to make this real.
It feels good to see the work out in the world.
wsj.com/cio-journal/mercedes…
WildEdge retweeted
Here is a complete family of open-weight, small models for text, vision, audio, and structured decisions.
They are around 2B parameters.
You can easily fine-tune these models and start using them in your application.
Introducing 𝙻𝙵𝙼𝟹: our new generation of Liquid Foundation Models. With a new tri-hybrid architecture and new multimodal encoders, the LFM3 family is the ideal model suite for specialized intelligence.
> 6 model instances, across text, audio, vision, and omni modalities, generative and decision models.
> 𝙻𝙵𝙼𝟹-𝟷.𝟽𝙱: text, 256k context window.
> 𝙻𝙵𝙼𝟹-𝚅𝙻-𝟸𝙱: vision-language model with video-input support for spatial understanding
> 𝙻𝙵𝙼𝟹-𝙰𝚞𝚍𝚒𝚘-𝟸𝙱: tool calls, voice chat, reasoning in 18 languages.
> 𝙻𝙵𝙼𝟹-𝙾𝚖𝚗𝚒-𝟸𝙱: an omni model combines vision, audio, and language.
> 𝚍𝟷-𝟸𝙱 and 𝚍𝟷-𝙾𝚖𝚗𝚒-𝟸𝙱: decision models with typed outputs and calibrated probabilities in a single pass. Omni adds audio input.
> fine-tune on 1 GPU.
> a new tri-hybrid architecture
> architecture optimized for 14 processor classes across GPUs, CPUs, and NPUs.
> 𝙻𝙵𝙼𝟹-𝚅𝙻-𝟸𝙱 is a beast. It performs competitively with Gemma 4 12B (6x larger), and Gemini 3.5 Flash Lite, and GPT-5.6 Luna, with 1.9B parameters.
> 𝙻𝙵𝙼𝟹-𝟷.𝟽𝙱 is #1 on the Pareto frontier of device intelligence 𝚙𝚒𝚙𝚎𝚝𝚝𝚎.𝚕𝚒𝚚𝚞𝚒𝚍.𝚊𝚒
Every modality for specialized intelligence. Pick the one that fits your use case, customize it, and run it on-device.
1/ 🧵
WildEdge retweeted
🚀Excited to announce Nanbeige4.3-33B!
To push the performance ceiling of small agents, our strategy in Nanbeige4.3 starts with building a powerful teacher first.
Nanbeige4.3-33B delivers SOTA agentic performance among open models at ~30B parameters:
* Strong results on agentic coding and cowork benchmarks.
* When integrated into mainstream agent harnesses, it tackles real-world tasks with cost-efficient inference.
* It will serve as a teacher for upcoming smaller Nanbeige models, starting with Nanbeige4.3-3B.
huggingface.co/Nanbeige/Nanb…
huggingface.co/Nanbeige/Nanb…
#LLM #Agent
WildEdge retweeted
Do you have 2 to 4 GB of vram?
Great! Now you can run a lil guy
Today, we’re releasing Bonsai 2 2B, 4B, and 9B.
Based on Qwen3.5, these new ternary models bring stronger reasoning, visual understanding, and tool calling to compact footprints designed for everyday devices:
• 2B: 0.43GB
• 4B: 0.95GB
• 9B: 2.0GB
For this release, we focused on what makes small models useful as local assistants: following instructions, avoiding repetitive loops, calling tools on mobile devices, and giving useful answers without unnecessary output. The 9B model uses roughly 40% fewer tokens on standard instruction-following and everyday math tasks. Solving the Riemann hypothesis can wait.
These models are successors to our first-generation Bonsai 1.7B, 4B, and 8B, and provide a foundation for bringing practical agentic functionality to local applications on personal computers, phones, and wearables.
The Bonsai 2 family now includes four sizes spanning 2B to 27B parameters. Developers can choose the balance of capability, memory use, and responsiveness that fits their application and hardware.
Available today under Apache 2.0.
WildEdge retweeted
What if Spotlight actually understood the meaning inside your files?
DigUp is an open-source macOS menu bar app powered by EmbeddingGemma 2 that brings 100% local, multimodal semantic search to your Mac.
Instead of searching for file names, you can use natural language to:
📸 Find specific scenes in your photos
⏳ Jump to the exact timestamp in audio or video
💻 Search through your codebase by intent
Everything runs locally in a single vector space.
WildEdge retweeted
I am not questioning if this is true (because it is), I am just wondering about how many iterations of innovation this will have to force on phone manufacturers in a short span of time
Qualcomm CEO reveals AI companies want phones running 100-billion-parameter models continuously by 2028--and confirms some are building phones
Cristiano Amon: “I have to be careful not to disclose names, but some AI companies—some of the companies at the forefront of AI—are telling Qualcomm, ‘By 2028, I need to make sure that I have the ability to run at least a 100-billion-parameter model in a phone, and I need to have this thing running kind of all the time.’”
Alex Heath: “So you’re saying there are AI companies that are talking about—”
Cristiano Amon: “Well, AI companies are starting to do phones.”
WildEdge retweeted
Qualcomm’s CEO on the coming AI phone supercycle
@cristianoamon on the China vs. US AI sentiment, how AI is changing smartphones and wearables, Qualcomm’s data center strategy, and 6G.
We discuss why he thinks phones aren’t going away, how agents will use our personal context, and why he expects smart glasses to become the biggest category of AI wearables. We also cover differences in consumer AI adoption between China and the US, Qualcomm’s acquisition of Modular to offer developers an alternative to Nvidia’s software, and what 6G could change about the devices we use.
This conversation was recorded live at Qualcomm’s recent Snapdragon Summit in Hawaii.
Timestamps:
00:18 How China and the US see AI
04:33 Should AI development slow down?
07:25 Trump, China, and chip geopolitics
08:25 Why the smartphone market is shrinking
11:15 Memory prices and pent-up demand
13:15 Can cloud agents replace the phone?
18:30 What is an AI smartphone?
25:05 The state of VR and smart glasses
29:30 Is there an iPod moment for wearables?
30:45 Why Qualcomm is building data centers
34:45 Modular and the fight against CUDA lock-in
39:05 What 5G actually changed
42:00 Why 6G is built for AI
45:05 When 6G arrives
Spotify: open.spotify.com/episode/3fT…
YouTube: youtu.be/FcYL3h5bwlk
Substack: sources.news/p/qualcomm-ceo-…
Apple Podcasts: podcasts.apple.com/us/podcas…
Thanks to the show’s premiere sponsors: @Atlassian, @meetgranola, and @mercury.
WildEdge retweeted
Thanks Google, Embedding Gemma 2 is a big deal 🫶
Multimodal embeddings (text, images, video, audio) can run on-device in the browser. No server, no API key, ~20–70 ms per query on WebGPU. Everything stays on your machine (ofc).
Now go build (new) things with it 🚀
Introducing EmbeddingGemma 2, a new open multimodal model that sets the standard for on-device efficiency.
- our first open, natively multimodal embedding model
- handles text, code, image, video, and audio tasks within a lightweight, modular 740M parameter form factor
- ideal for offline, privacy-first RAG when paired with Gemma 4
- outperforms some specialist models more than twice its size
Weights available now on Hugging Face.
WildEdge retweeted
Agents should be able to report what they couldn't do. We just shipped submit_feedback in the @getWildEdge MCP server.
"another agent drafts the PR" means any caller writes the prompt with repo access. That's prompt injection with extra steps.
Human has to make the final call.
WildEdge retweeted
Let's go local AI!
At a time when everyone is complaining about shortages of energy and compute for AI & costs of tokens, we need on-device AI more than ever! In addition, it's virtually free & fully privacy preserving!
Today we're launching Desert Ant Labs: a European frontier AI lab building on-device intelligence.
18 models across audio, vision, and text. SDKs for Swift, Kotlin, and JavaScript.
No tokens. No logins. Nothing leaves the device.
desertant.com
WildEdge retweeted
Today we're launching Desert Ant Labs: a European frontier AI lab building on-device intelligence.
18 models across audio, vision, and text. SDKs for Swift, Kotlin, and JavaScript.
No tokens. No logins. Nothing leaves the device.
desertant.com
WildEdge retweeted
Tiny models are becoming software primitives
This 27.5KB model runs locally in your browser, syntax-highlighting code without being told the language
I trained a small model to do syntax highlighting in the browser with GPU.
Meet gpu-lexer from Vercel Labs: Small (27.5KB), fast (runs on WebGPU), and language-agnostic (model guesses the syntax).
gpu-lexer.vercel.app
It is experimental and built for learning!
WildEdge retweeted
Today we’re open-sourcing Lily, the local inference engine we built for hybrid compute in Perplexity Computer.
Lily is specialized for Qwen3.6-35B-A3B on Apple silicon, built so on-device compute doesn’t bottleneck Computer tasks.
Read more: perplexity.ai/hub/blog/optim…
WildEdge retweeted
i'm amazed. a 3B vision model tracked one spot over four visits. no cloud saw a photo. a 4-bit build runs on my iphone.
i can't stop thinking about the 4.6 billion people without access to essential health services. it fits in their pocket today.
how many lives could that save?
A single skin photo can show you a spot. It cannot show you what changed.
We gave LFM2.5-VL-3B from @LiquidAI four dated views of the same generated surface. It coordinated a local tool pipeline to follow one region through time and surface the comparison for clinician review.
WildEdge retweeted
Best-in-class PII Detection...running entirely on your device 🤯
perplexity.ai/hub/blog/pii-t…
WildEdge retweeted
Super impressive!!
We’ve been arguing for small models for specialized tasks for a year! It’s time for domain specific superintelligence 💪🏽
Paper 1: arxiv.org/pdf/2507.13966
Paper 2: arxiv.org/pdf/2601.15160
WildEdge retweeted
A single skin photo can show you a spot. It cannot show you what changed.
We gave LFM2.5-VL-3B from @LiquidAI four dated views of the same generated surface. It coordinated a local tool pipeline to follow one region through time and surface the comparison for clinician review.
WildEdge retweeted
So many vendors are NOT getting this
I have one or two agents I use and like. For anyone else: give me an MCP interface to connect these agents to so I can use your service
Unless your a frontier AI lab, I prob don't want to use your agent, sorry
WildEdge retweeted
Implementing a car recognition app used to mean collecting images, labeling them, and training a model before you could ship anything. Now you send a photo to an LMM, and get back the results.
Check out @getWildEdge's blog post on reasons to train a custom classifier.
WildEdge retweeted
DROP EVERYTHING
The bible for running LLMs locally is now available online to read for free
Covers what to use on
- Laptop / edge / odd hardware
- Mac-first workflows
- Single RTX GPUs
- 2-4+ NVIDIA / CUDA GPUs
- General production serving
- Long-context / MoE / routing
- NVIDIA max performance
- Cluster orchestration
Software
- llama.cpp
- MLX / MLX-LM
- ExLlamaV2
- ExLlamaV3
- vLLM
- SGLang
- TensorRT-LLM
- NVIDIA Dynamo
You should read this, and if you cannot now then you most definitely wanna bookmark it for later
Local AI FTW
