The Eachlabs Index: From Models to Systems 815% Growth in 10 Months

Hey Eachlabbers, Eftal here.I’ve been going through our October analytics report, and to be honest this one feels different. You can almost feel the shift happening under the surface. The headline numbers are already insane: an 815% transaction volume growth since January.

1,742 active apps/ projects in October. 

And Text-to-Text models up 4,686%. Some workflows are using these models alot.

But beyond the wild metrics, October wasn’t just about growth. It was about modularity finally becoming mainstream the month when developers stopped just connecting models and started composing systems.

When Text Models Fought Back

Let’s start with something that surprised even us: Text-to-Text workflows are back in the spotlight.

After months of image and video dominance, October flipped the script. Developers started chaining LLMs again not for simple prompts, but for orchestration: summarization + reasoning + reformatting inside full pipelines.

The 4,686% growth isn’t just a number it’s a sign that text models found a second life as glue. They’re now the routers, the interpreters, the logic engines behind multimodal stacks. Image-to-Image may have dominated the creative side, but Text-to-Text quietly became the nervous system.

It’s funny back in May we thought visual AI would overshadow text forever. Turns out, once everything became multimodal, text found a new purpose as infrastructure.

The Modular Moment

If you’re wondering where the real action happened, it's in the way people are Eachlabbing* now.

October’s report shows 49 projects doing 10× Eachlabbing, orchestrating over 20 models each, and 156 projects doing 2×. That’s not experimentation, that's architecture.

I’ve been talking to teams who are literally treating models as microservices. They’re swapping diffusion for transformer blocks, mixing audio synthesis with image-to-video, even chaining reasoning models with stylizers. Two years ago this was an academic dream. Now it’s running on production APIs.

And those 1,742 active apps? Sure, not all of them are massive, but the pattern is clear: devs aren’t just testing models anymore, they're building modular AI systems.

Eachlabs didn’t invent the idea, but we might be the first platform where it actually works at scale.

*Eachlabbing: If Eachlabbers (who are using our system in depth with workflows and complex designs) are using +1 model for their projects we call this Eachlabbing. 

The Real Cost of Curation

Here’s something we rarely talk about: the brutal selection process behind those model numbers.

In October alone, we tested 152 new models, added 57, and delisted 31. That means two out of three don’t make it past internal testing.

On paper, we now have 338 models actively used in the last six months but what you don’t see is the countless others that never reached production. 

We reject more than we accept because stability, latency, and price-performance ratios matter more than catalog size.

Model curation isn’t glamorous workit’s weeks of testing, failed runs, and fine-tuning batch limits but it’s what makes Eachlabs feel consistent even when the ecosystem is chaotic.

Volume Explosion, Real Adoption

That 815% transaction volume growth since January could be mistaken for marketing fluff, but it’s very real. The difference this time is composition depth.

Last quarter’s volume came mostly from single-model use cases. October’s spike is driven by chained workflows: image-to-video + caption + upscale + audio + caption running thousands of times a day.

The Image-to-Image category is still growing fast (+2,525%), and Text-to-Image held its place at +590%. But the story isn’t just about which vertical wins, it's about the shift from “AI features” to “AI infrastructure.”

Developers are building pipelines that look more like cloud architectures than creative tools.

When you can trace 80% of platform usage back to fewer than 50 well-architected systems, you realize this isn’t hype anymore. It’s production.

The Bigger Picture

October was the month AI went modular.

And honestly, I wouldn’t want to be anywhere else while it’s happening.

Eachlabbers are trying/ using the combination of different models for the best output. 

As Eachlabs team we focused more on ready to use workflows, serving some trends and helping you all to reach your dream outputs.

Eftal