Twitxy Saved

Quantinine

Verified account This account is verified. Learn more

@quantinine

Joined October 2021
134 Following 12K Followers 1.2K Posts

Latest posts 5

Quantinine Verified account @quantinine
Kolibri-1 DE is now open 🇩🇪

Aleph Alpha’s Kolibri-1, further trained for German compliance work. Apache 2.0:
huggingface.co/Quantinine/Kol…

On 4,272 German test documents, without reasoning:
→ score 41.1 → 76.5 (out of 100)
→ correct documents recognised as correct: 58.7% → 88.8%

@IlhanScheer @tfburns @letiepi @alessio__serra @resiliem @tradeqvest @Aleph__Alpha
huggingface.co Quantinine/Kolibri-1-DE-NVFP4A16 · Hugging Face Text Generation model with 40B parameters. Download, explore files and run it on Hugging Face.
Quantinine Verified account @quantinine Oct 23
Kolibri-1 DE is here 🇩🇪

Aleph Alpha's Kolibri-1 + German compliance training

Access coming mid-November.
Hosted in Germany.

DM for early access.

@IlhanScheer @tfburns @letiepi @alessio__serra @resiliem @tradeqvest @Aleph__Alpha
View quoted post on X
Quantinine Verified account @quantinine
This is actually an insane find
Cohere Verified account @cohere Oct 26
While evaluating Qwen3.8-Flash-Next in a controlled sandbox environment, Cohere's research team observed something unexpected.

Asked to generate a defensive cybersecurity detection rule, the model instead inventoried aspects of the machine it was running on and attempted to upload a probe file to an Alibaba Cloud storage bucket in Beijing. Notably, the model's reasoning trace stated that it would not invoke any tools before taking these actions.

Cohere's sandbox blocked the attempt, and no external communication succeeded. We have not determined the root cause, and we are not asserting malicious intent.

But the incident raises important questions for organizations deploying agentic AI systems: Can models take actions that differ from their stated reasoning, and how should security teams evaluate trust, provenance, and behavior in open-weight models?

We are publishing our findings because organizations making model adoption decisions deserve as much information as possible. We hope this contributes to greater transparency, stronger security practices, and continued collaboration across the AI ecosystem.

Read the full report: cohere.com/blog/concernin…
View quoted post on X
Quantinine Verified account @quantinine
Kolibri-1, now 42% lighter ⚡

Releasing Kolibri-1-NVFP4A16-FP8: Aleph Alpha's Kolibri-1 in 4-bit, on par with the original across 10,692 agentic episodes and 24 benchmarks.

huggingface.co/Quantinine/Kol…

@IlhanScheer @tfburns @letiepi @alessio__serra @resiliem @tradeqvest
huggingface.co Quantinine/Kolibri-1-NVFP4A16-FP8 · Hugging Face Text Generation model with 40B parameters. Download, explore files and run it on Hugging Face.
Aleph Alpha @Aleph__Alpha Oct 3
Small bird, fast wings, Kolibri is here.

78B parameters. 3.46B active. Up to 1M tokens of context. Built in Europe.

Now the weights are yours. Run it on your own hardware, under Apache 2.0.
View quoted post on X