Author: Jiri Krecek
-
AI@Home – Granite 3B MoE
Model conditions & context So, today I’m looking at IBM’s models again. This time it is one of their tiny series, but unlike the 2B dense model I tested last week, this is a Mixture of Experts model. And that always gets my attention (no pun intended). Let’s dive right in. I picked a very…
-
AI@Home: IBM Granite3 2B
Model conditions & context Today we are looking at another model from the family of lighter weights — and this time it is from IBM. Should be interesting. clocking at a puny 2B this Granite 3 2B model is one of the lightest, if not the lightest, I’ve tested. Let’s dive in to see how…
-
AI@Home – Ministral 3 3B
Model conditions & context Today we are looking at the smallest member of Mistral’s 3rd generation family — the Ministral 3B. As tiny as it is, it’s weight is actually close to 4B. Now, despite the shortcomings and propensity to use bullet points and to be constantly of service and trying to be helpful, Mistral…
-
Nemotron 3 Nano
Conditions & Context Today we have a doozy. A relatively tiny model with a really big heart. I just recently tested Nvidia’s Nemotron Mini, which is a puny 4B model and it performed alright. I’ve read up on their Nemotron 3 Nano, which is slightly heavier – sitting right above 5 GB in weight, but…
-
AI@Home – Nemotron Mini 4B
Conditions & Context I’m hungry. Today we are tasting a not-so-fresh appetizer — NVIDIA’s Nemotron Mini. And running a 4B model on a fast card may give us some good speeds. I dare say we may be up for a treat. Did it live up to already low expectations? Let’s dive right in and see…
-
AI@Home – Cogito V1 14B Review
Conditions & Context After doing a review of its little 8B brother a couple days ago, today we are looking at Cogito V1 14B model and I’m curious how it would fare in my very simple test. Unlike the 8B, which was based on Meta’s Llama model, this 14B variant is forked off open source…
-
AI@Home – Cogito V1 8B Review
Conditions & Context Today I’m looking at Cogito V1 8B model in Q4 K M quantization. This is Meta’s Llama 3.2 under the hood, but with Cogito’s proprietary self-improving IDA training loop baked in. Is it better than Llama? Let’s dive in and see. I picked a very simple prompt which contains a mixture of…
-
AI@Home – Apriel 1.6 15B Thinker Review
Conditions & Context Today we have something genuinely unusual on the bench. Not Meta, not Google, not Mistral. This one comes from ServiceNow — yes, the enterprise workflow automation company. I still remember using their IT ticketing system at my previous company. They know a thing or two about tech, so should be interesting… Their…
-
AI@Home – Mistral 3.1 Small 24B Review
Conditions & Context Today we are going back to France! On the table is a Mistral 3.1 Small with a decent 24B weight. It will be a tight squeeze onto a 16GB GPU, so I expect some CPU cores being lit up, but let’s see if it did as bad as Gwen. Let’s dive right…