Everyone knows about large language models: piles of matrices that flop around on power-hungry GPU datacenters and cost a small fortune to depend on.

But what's going on in the small end? What sorts of language tasks can you solve using only hardware that you already own?

This talk is about the state of the art for small language models in 2026. I'll compare what can be done with models small enough to run on verious types of hardware: a smartwatch, a raspberry pi, a smartphone or a laptop.

If you're curious about language models, but also care about software freedom, privacy, and reducing power consumption, this talk is for you.

---

Speakers for Small Language Models: What Can They Actually Do in 2026?:

---

Metadata for Small Language Models: What Can They Actually Do in 2026?

To be recorded: Yes
To be streamed: Yes

---

URLs for Small Language Models: What Can They Actually Do in 2026?

No URLs found.

---

Schedule for Small Language Models: What Can They Actually Do in 2026?

• Friday, Jul 17th, 2026, 16:00 (CEST) - Friday, Jul 17th, 2026, 17:00 (CEST) at Speakers Tent