The flower photo by Sergey Shmidt on Unsplash

İvmeLabs

Welcome to İvmeLabs! We are a Turkish HuggingFace organization that builds AI models that are small and push above their weight.

Our main proposition for our language models is that they're small, but they do one thing (or multiple things if we are crazy) right and do it well.

Our Model Families

İvme

İvme-Conversate (Codenamed Apple): İvme-Conversate is our series of language models designed for prediction and conversation using human language.

İvme-Coder (Codenamed Otter): İvme-Coder is our first series designed around a singular category. The İvme-Coder series focuses on trying to write code at incredibly small sizes.

İvme-Triage (Upcoming, Codenamed Sheepdog): İvme-Triage is our series of language models designed to route prompts to the correct model in a given set of language models with their metrics.

İvme-Classify (Upcoming, Codenamed Orange): İvme-Classify is our series of language models designed for zero-shot classification at small sizes.

İvme-Polisher (Upcoming): İvme-Polisher is our series of diffusion language models (That might change; we don't know) designed for polishing user text on the go, such as fixing grammar or adding markdown styling.

İvme-Corrector (Upcoming): İvme-Corrector is our series of language models designed for autocorrect tasks and word prediction specifically for Turkish and its morphology.

İvme-Speak (Upcoming, Codenamed Blueberry): İvme-Speak is our series of TTS models designed to give realistic-sounding voices at small parameter sizes. Engineered for assistants and the like, where you will be hearing it constantly.

İvme-Ultra (Upcoming, Codenamed Belladonna): İvme-Ultra is our planned first model series with MoE. The first model is going to be 1.2B, which is not really small, but we are going to give it a weird quirk which might compensate for that. We don't want to give more details about it, as it's early, and we don't want to leak the ideas on it.

Variants

Conversate:

The Conversate family is going through a variant reformation. The new variant sizings will be like this:

Conversate-U: 500M~ parameters range.

Conversate-XXL: 150M to 250M parameters range.

Conversate-XL: 100M to 150M parameters range.

Conversate-L: 60M to 90M parameters range.

Conversate-M: 30M to 50M parameters range.

Conversate: 20M to 30M parameters range.

Conversate-S: 20M to 10M parameters range.

Conversate-N: 10M and lower.

There's a lot, but for a long time I'd say don't expect XXL and U, and probably L, S, and N due to them being mostly ineffective ranges. Just consider a bunch of these as name reservations.

Expİvme

Expİvme is our series of models mainly made for experimentation. They lack codenames, and they're not included here. Expİvme models are mostly not expected to perform in benchmarks and instead try to see how different architectures or training styles work on small models. When an Expİvme family develops well enough, it can be promoted to the normal İvme family, marking it as ready for production.

Note the Budget

We will TRY to bring models on time, but we CAN'T really verify that they'll come as soon as we say they are coming. We hit budget problems and noisy hosts. Even backing up İvme-Conversate-v1-Base's training data was hard for us after our host refused to boot. But don't worry, we will try our best as long as our host won't fail and our budget won't hit its limit!

Current Status

Last Updated: August 28, 2026

Ivme-Conversate-XL-v1.1-Base has also started to show abnormal signs. Ok I am going to be honest here: just stop using Conversate-XL-v1 and Conversate-XL-v1.1; they're just too broken to be fixed. I already removed v1.1 from the leaderboards, and I am going to consider both of them as fundamentally broken. Also, being honest, I stopped seeing any point in making the Conversate line like this if it's just another generic SLM. So after I hopefully release Conversate-XL-v2, releases will be more infrequent but more sophisticated for Conversate models. Also, I will divert the resources to other models for a while.