We just launched the @NeuraLake Inference API 🚀
It looks simple. That's the point.
Agents don't need thousands of models to choose from. They need inference that just works. Fast, reliable, and efficient.
https://t.co/XyIlqircKN
🚀 NeuraLake receives international recognition at WCSA 2025
We’ve been awarded the WCSA2025 Special Distinction by the CBC AI & Digitalisation Hub, granted by the World CleanTech StartUPs Awards, in partnership with CleanTech Business Club and @Acwa_en Power.
NeuraLake remains focused on empowering companies of all sizes and large industrial players to build specialized, private, and operational AI — on-premise and especially at the edge, where data is generated and real-time decisions happen.
We know you’re waiting for iSA-03-Edge and IoT on @huggingface
But the future arrived faster than we planned 🚀
An unplanned trip shifted everything overnight 😅
What’s coming is bigger than we expected
The roadmap is being rewritten right now and only thing we can say is👀
We get it—the name’s a mouthful!
Just regrouped with the team and here’s the call:
We’ve dropped the non-hybrid variants. From now on, meet the lineup:
• iSA-03-Nano (1B)
• iSA-03-Mini (3B) ← now live
• iSA-03-Edge (7B)
• iSA-03-Max (???B) — little surprise
All hybrid. All iSA-03.
The non-hybrids felt too close to base models in early tests — not worth it.
We extended the context length for long-form reasoning when needed.
And yeah… we made the name less dumb and a lot simpler.
Only Mini is available for now — more soon.
#AI #NeuraLake
Everyone’s building AI agents with giant LLMs.
But NVIDIA says that’s wasteful. And we agree.
Small Language Models (SLMs) are faster, cheaper, and better aligned.
The real bottleneck? Data.
That’s where synthetic data steps in. That's where we step in.
Small Language Models are the Future of Agentic AI
Lots to gain from building agentic systems with small language models.
Capabilities are increasing rapidly!
AI devs should be exploring SLMs.
Here are my notes: