Switching models in the middle of a long session sounds risky: what if the history gets wiped? In Verboo Code it doesn't. The /model command only swaps the engine answering you, never the conversation.
Why doesn't switching models mid-session erase the context?
Because the code only touches one isolated field of the session state, mainLoopModel, and never touches the message history. Verified directly in src/commands/model/model.tsx in the verbeux-ai/code repository: both the direct command and the interactive menu end up in the same function, which only swaps mainLoopModel in AppState. The next reply already comes from the new model, with everything said before still visible to it.
How do I switch models directly, without opening any menu?
Pass the model ID together with the command. The switch is immediate, no confirmation needed:
/model glm-5.2Real output from the command:
Modelo alterado para glm-5.2.That confirmation prints in Portuguese regardless of your system locale. Confirmed in the source: Verboo Code's model command hardcodes pt-BR strings for this flow, so every account sees the same message. If the ID doesn't exist on your plan, Verboo Code refuses and points you to the right path, also in Portuguese:
O modelo 'gpt-99' não está disponível. Execute /model para escolher um modelo liberado.How do I see every model unlocked on my plan?
Run /model with no argument. It opens a menu with the models your account has already unlocked, fetched live via API (function loadUnlockedModels, same file):
/modelVerboo Code currently unlocks 13 open models, all with a 1M context window except one:
| Model | Context |
|---|---|
glm-5.2 | 1M |
glm-5.3 | 1M |
glm-5.3-flash | 1M |
deepseek-v4-flash | 1M |
deepseek-v4-flash-0731 | 1M |
deepseek-v4-pro | 1M |
deepseek-v4-pro-0813 | 1M |
deepseek-v4.1-flash | 1M |
kimi-k2.7 | 1M |
mimo-v2.5 | 1M |
mimo-v2.5-pro | 1M |
minimax-m3 | 1M |
qwen3.8-27b | 262K |
List checked on verboo.ai on 2026-09-14. Each plan unlocks a subset of these 13; the /model menu only shows what your account already has access to. The pay-per-token pricing page lists a separate, smaller catalog for usage-based billing instead of a subscription.
How do I refresh the catalog if a new model isn't showing up?
The /model refresh command fetches the list again, skipping the local cache:
/model refreshCatálogo atualizado: 13 modelo(s).
What else changes when you switch models mid-task?
Three things, all automatic, confirmed by reading clearRouterRateLimitIfModelChanged and the rest of model.tsx:
- The previous model's rate limit clears. If you hit the limit on one specific model, switching to another frees your account instantly, without waiting for the reset window.
- Fast mode adjusts itself. If the new model doesn't support fast mode and it was on, Verboo Code turns it off and tells you (
Fast mode OFF). If it does and fast mode was available, it turns back on (Fast mode ON). - Billing category can change. Some models count as extra usage on your plan; when that's the case, the switch message itself warns you with
Billed as extra usage.
Gotcha: separate from the default model, Verboo Code keeps a value scoped just to the current session (mainLoopModelForSession). Running /model current shows when it exists, with the label (session override from plan mode) next to the model name. The detail that trips people up: any manual model switch, whether with /model <id> or through the menu, clears that value right away. There's no way to keep the override and only swap the base model at the same time.
Switching models mid-task usually comes from a real reason: a limit you hit, a bad answer, a task that calls for different reasoning. In Verboo Code that switch never runs into counted tokens, because plans run with unlimited tokens, and the entire session history follows you from one model to the next.
