Borg Fusion Profiles

What Borg Fusion does

Borg Fusion is a routing meta-model, not a separately trained foundation model. It selects an underlying model for each step of a task instead of committing the whole session to one provider. Its general profile is useful for general tasks. The underlying model may change between steps or sessions, so do not infer a fixed model's capabilities, context window, or token price from the routing label.

Choose a profile

Borg Fusion is the general routing option. Pick it when you want Borg to manage model selection across general work.

Borg Fusion Coding is tuned for coding tasks while still selecting an underlying model for each step. Choose it for coding work when you prefer routing over manually selecting a vendor lane.

Borg Fusion Design is tuned for design tasks and selects an underlying model for each step. Use it for design work when you want Borg to manage the choice.

Borg Fusion Fast is tuned for speed and is useful for time-sensitive tasks. It still chooses an underlying model for each step, rather than promising one fixed model.

Atlas Max is Borg's high-capacity lane for work that benefits from a larger execution budget. It is intended for demanding, high-capacity tasks. Its label does not identify a fixed foundation model or a guaranteed context size.

Profile or direct lane?

Choose a profile when the task category or need is clear, but you want Borg to choose an underlying model. Choose a direct lane when you need a fixed model choice, such as when testing a prompt against one model or keeping behavior consistent across a workflow.

A profile is not a claim that every step uses the same underlying model. If a particular capability or limit matters, look at the selected model's own information. For repeatable model-specific work, use the direct lane ID listed on that model's page. For profile IDs, copy the exact value from the catalog rather than deriving it from the display name.

Profiles

Borg Fusion

Borg Fusion is one of Borg’s routing meta-models, not a separately trained foundation model. Fusion selects an underlying model for each step of a task rather than committing the entire session to one provider.

Lane IDs: borg-fusion, atlas-fusion. Thinking modes: low, medium, high.

Borg Fusion Coding

Borg Fusion Coding is one of Borg’s routing meta-models, not a separately trained foundation model. The Fusion Coding profile is tuned for coding tasks while still selecting an underlying model for each step.

Lane IDs: borg-fusion-coding, atlas-fusion-coding. Thinking modes: low, medium, high, max.

Borg Fusion Design

Borg Fusion Design is one of Borg’s routing meta-models, not a separately trained foundation model. The Fusion Design profile is tuned for design tasks while still selecting an underlying model for each step.

Lane IDs: borg-fusion-design, atlas-fusion-design. Thinking modes: low, medium, high, max.

Borg Fusion Fast

Borg Fusion Fast is one of Borg’s routing meta-models, not a separately trained foundation model. The Fusion Fast profile is tuned for speed while still selecting an underlying model for each step.

Lane IDs: borg-fusion-fast, atlas-fusion-fast. Thinking modes: low, medium, high, max.

Atlas Max

Atlas Max is one of Borg’s routing meta-models, not a separately trained foundation model. Atlas Max is Borg’s high-capacity lane for work that benefits from a larger execution budget.

Lane IDs: atlas-max. Thinking modes: Not published.

Frequently asked questions

Is Borg Fusion a foundation model?

No. It is a routing meta-model that selects an underlying model for each task step.

Which profile is tuned for coding?

Borg Fusion Coding is tuned for coding tasks.

When should I choose a direct lane?

Choose one when you want to keep a workflow on a particular model.