# Open models only: back to school challenge

> Source: <https://wagtail.org/blog/open-models-only-adoption-challenge/>
> Published: 2026-09-03 00:00:00+00:00

# Open models only: back to school challenge

## Ditch the subscription, spend the whole of September on open weight models

If you’ve been [comparing open weight AI models and providers](/blog/comparing-open-weight-ai-models-and-providers/) like us, now’s the time for action, in the form of a challenge. This month, **ditch your AI coding subscription and try to spend the whole of September using only the one efficient open weight model**: [GLM 5.3 Flash by Z.ai](https://z.ai/blog/glm-5.3-flash). Here’s how, and what to expect.

## Why open weight, why GLM 5.3 Flash

The overall strategy is best explained in Mozilla’s [Owners, not renters AI strategy](https://blog.mozilla.org/en/mozilla/mozilla-open-source-ai-strategy/). The TL;DR; is you get greater flexibility, more transparency on your usage, and these days pretty much the exact same capabilities.

### New efficient models

For agentic coding specifically, open weight models are now so capable that the bottleneck on your results is probably either AI cost or your own expertise. GLM 5.3 Flash specifically is at the [very top of the pareto frontier](https://pareto-eco-sum.netlify.app/), with benchmark scores at the same level as much heavier models:

Those excellent scores on the [Artificial Analysis index](https://artificialanalysis.ai/) combine with excellent usability:

- 1M context window, so any extended coding task is possible.
- Vision support, so it can build off screenshots or QA visually.
- Available across a wide range of providers.

And a very low price which reflects how efficient it is to host this model, based on its size and architecture.

## How to switch

We’ve compiled our own opinionated [agentic engineering recommendations](/ai/agentic-engineering-recommendations/), for people who want to follow our [AI guiding principles](/ai/ai-guiding-principles/).

If you’re not ready to fully switch just yet, complement your existing usage with:

[Open Code Review](https://open-codereview.ai/)for AI code reviews.[AgentsView](https://www.agentsview.io/)for usage tracking.

Both tools work great with existing wokflows. Here’s an example of AgentsView reporting on models in use over 30 days, showing DeepSeek V4 Flash usage across a wide range of tools:

Those tools are exactly what we’ll be using when trialing Wagtail’s [Developer Experience with agent skills](https://github.com/wagtail/wagtail/discussions/14438), so it’s crucial we can come forward with strong recommendations for our community.