Skip to main content
Forge is Orkestral’s bundled local code model. It runs entirely on your machine (Qwen2.5-Coder via node-llama-cpp, GGUF) and executes the actual code changes, so routine work costs nothing in API fees.

​
The idea: premium plans, local executes

A premium model is great at planning. The Forge is great (and free) at doing. Orkestral splits the job so you pay only for the thinking, not the typing.

​
Fast Apply

The Forge pairs with Orkestral’s own Fast Apply, a deterministic engine that merges edits into your files:
1

Exact match

Applies the change where it matches exactly.
2

Whitespace-normalized

Retries ignoring indentation/whitespace differences.
3

Safe fuzzy

A single-match fuzzy pass, and it rejects anything ambiguous rather than writing the wrong content.
Fast Apply only touches the changed lines, it never rewrites the whole file, and it needs no external service.

​
$0 by design

Offline and free

Because the model is bundled and runs locally, executing changes uses no API credits and works with no internet connection. A cost view shows how many runs were resolved locally vs. escalated to a premium model.

​
When it escalates

If the Forge can’t apply a change confidently, Orkestral escalates once to a premium model as a fallback, so correctness is never sacrificed for cost.
The Forge model ships inside the installer, so it’s ready on first launch with no extra download.