Choose a model

Switch models in the "Model" submenu of the composer settings. Every row ends with a reference multiplier: roughly what the model costs on a typical task relative to ClawsGO 1.5 Silicon (x1.0). Actual charges are metered per token at each model's own input, output and cache rates, see Credits.

ClawsGO 1.5

ClawsGO 1.5 is our own model line, built on leading open models, post-trained and fine-tuned for the ClawsGO Science Harness. The three tiers trade depth against cost:

New accounts start on Lithium.

Third-party models

The menu also carries Anthropic and OpenAI models. The ClawsGO Science Harness stays the same; only the underlying model changes:

"More → View all models" at the bottom of the menu opens a searchable dialog grouped by vendor, with Claude Fable 5.1 and previous-generation models. Switching to the Claude Fable line asks for confirmation and shows its credit usage relative to Claude Opus 5.5.

Models outside your plan are greyed out; clicking one opens the subscription section. Each plan's model range is listed in Plans.

Which to pick

• Everyday questions, formatting, small code changes: Lithium.
• Multi-paper reviews, data analysis, drafting: Silicon or Claude Sonnet 5.5.
• Long, complex research that needs repeated self-checking: Titanium, Claude Opus 5.5 or GPT 6 Astra.
• The hardest problems: Claude Fable 5.1.

Switching mid-task

Switching on the new-task page has no side effects. Switching inside an existing task asks for confirmation: it resets the context cache, so the next turn costs more, then back to normal. The switch applies to later turns only; the turn already running is unaffected.

If a model declines your content, switch to another model and reply "Continue". Third-party models decline based on their vendor's safety policy, which ClawsGO can't override.

Subagent model

The "Subagents" submenu sets the model for subagents the main agent dispatches, including cluster members. It appears in Delegate and Cluster modes. The default is "Same as main model"; you can also pick another model.

• Every subagent uses this one model. You choose it; the main agent cannot change it.
• It only affects subagents: the main conversation, web page summaries and context compaction still use the main model.
• Availability, locked rows and the Claude Fable confirmation work the same as in the "Model" submenu.
• Switching the subagent model inside an existing task does not break the main conversation's context cache, so there is no confirmation.

Subagent calls are billed at the subagent model's rates. A common setup: a strong main model plans and reviews while subagents on Lithium handle bulk searching and file reading. Or the reverse, with a stronger model doing the key reviews.

Context window

The "Context" submenu has three settings:

• 256K: the standard window, right for most tasks. Earlier content is compacted automatically as the limit approaches.
• 512K: more room for long papers and multi-step work. There is no feature fee, but each turn sends more input, so credit usage rises. Plus and above.
• 1M: room for more papers, data and code, with fewer details lost. Each turn sends even more input, so credit usage rises significantly. Pro and above.

A new task starts with the window you last picked, or 256K if your plan doesn't include it. A task on 512K or 1M shows the window on the settings button, with 1M in amber. Switching to 1M asks for confirmation.

Reasoning

Five levels: Low, Medium, High, Extra High, Max. The default is High. Every mode runs at the level you pick. In Cluster mode, cluster members use it by default; when orchestrating, the main agent can raise or lower it for individual steps.