AI research and products that make smaller models worth building

Anthropic 2 is an independent research lab building compact models and the infrastructure to decide—honestly—when they beat the alternatives.

MODEL 001JULY 2026

Trained with Distillery

TinyRamp

Read the model report

One portable finance generalist. Distilled from a large API teacher into a 0.5B model.

0.5Bparameters

2finance tasks

1set of weights

Distillation product · available now

Distillery

From a dataset to a portable model in three lines.

Open Distillery

distill.py● ready

distillery = Distillery(api_key=os.environ["DISTILLERY_API_KEY"])
dataset = distillery.datasets.create("./finance_world.jsonl")
run = distillery.distill(dataset, recipe="auto").wait()

Latest releases

View all research

Model releaseJuly 2026

Introducing TinyRamp

A 0.5B finance generalist trained with Distillery and evaluated against its teacher.

ProductJuly 2026

Distillery

Turn an evaluation dataset into a smaller, portable model with one resolved recipe.

ResearchJuly 2026

When should a model be smaller?

A systems-and-economics evaluation of distillation as a deployment decision.

We build models that earn the right to be smaller.

ModelsTinyRamp is our first portable finance generalist.SystemsDistillery makes method selection and evaluation reproducible.EvaluationQuality, uncertainty, systems behavior, and economics share one release gate.