First-class Models
Xum ships with curated models kept up to date with the frontier. Use any custom model with/model <provider:model_id>.
Xum supports the GPT-6 tiers (GPT-6.2 Sol, GPT-6.2 Astra, and GPT-6 Luna); older OpenAI GPT models are no longer supported. GPT-6.1 Sol and GPT-6 Astra stay usable as the custom model strings
openai:gpt-6.1-sol and openai:gpt-6-astra.
GPT-5.6 models added from Settings → Models (gpt-5.6, gpt-5.6-sol, gpt-5.6-terra, and gpt-5.6-luna) support thinking from off through max and Pro reasoning mode. Other older GPT models stay unsupported.
GPT-6 Luna supports thinking from off through max on the default Responses API route. On the opt-in Chat Completions route, Xum disables its reasoning because OpenAI only supports function calling there with reasoning off.
GPT-6.2 Sol (the gpt and sol aliases), GPT-6.2 Astra (the astra alias), GPT-6.1 Sol, and GPT-6 Astra support thinking from low through max; OpenAI does not allow turning their reasoning off. Use the default Responses API route with them, because OpenAI does not support tool calling for these models on Chat Completions.
Codex subscription Fast mode
On a Codex OAuth model, open the Thinking selector and turn on Fast mode. You can also type/fast or use Toggle Fast Mode in the Command Palette.
Toggle Fast off to return to Standard for that chat. Xum still uses your Codex subscription.
Fast mode uses more of your plan’s allowance or credits than Standard.
OpenAI Ultrafast
Turn on Ultrafast mode in the Thinking selector or type/ultrafast. You can also use Toggle Ultrafast Mode in the Command Palette or use its keyboard shortcut. To set a default for new chats, select ultrafast in Settings → Providers → OpenAI.
Fast and Ultrafast are mutually exclusive. Turn either mode off to return to Standard. Xum saves your choice per chat, even when you switch models. Chats without a choice use the provider’s service tier.
Ultrafast supports GPT-6.2 Sol, GPT-6.2 Astra, GPT-6.1 Sol, and GPT-6 Astra. It requires the Responses API. Supported routes include direct OpenAI (API key or Codex OAuth), Coder instances of type OpenAI, and custom OpenAI Responses providers. Chat Completions, Coder OpenAI-compatible instances, OpenRouter, and GitHub Copilot do not support Ultrafast. On unsupported models or routes, Xum omits Ultrafast instead of switching to Fast.
With Codex OAuth, Ultrafast uses your Codex subscription. Codex plan, workspace, and residency requirements apply.
For API key requests, Xum estimates Ultrafast costs at 6× the Standard price. OpenAI charges 3× for GPT-6.2 Sol and Astra through February 28, 2027. Until then, Xum estimates twice the actual API cost for these models.
OpenAI does not offer the Flex tier for GPT-6.2 Sol and Astra. If Settings → Providers → OpenAI sets the service tier to flex, their requests run at Standard instead.
Pro reasoning mode
The GPT-6 tiers (GPT-6.2 Sol, GPT-6.2 Astra, GPT-6.1 Sol, GPT-6 Astra, and GPT-6 Luna) and the GPT-5.6 models added from Settings → Models support OpenAI’s pro reasoning mode. Open the reasoning selector next to the model picker and enable Pro mode (or run “Toggle Pro Reasoning Mode” from the Command Palette) to sendreasoning.mode: "pro" with each request. The setting is saved per workspace. Pro mode can consume more tokens and responses can take noticeably longer. Pro is available on supported direct OpenAI Responses API routes. Gateway routes, Chat Completions, and Codex OAuth use standard mode.
Model Selection
Keyboard shortcuts:- Cycle models
- macOS:
Cmd+/ - Windows/Linux:
Ctrl+/
- macOS:
Cmd+Shift+P / Ctrl+Shift+P):
- Type “model”
- Select “Change Model”
- Choose from available models
provider:model-name
Add custom models
Under Settings → Models, choose a provider and use the Model ID field to search its catalog or enter an ID. Clicking a suggestion, or selecting one with arrow keys and Enter, adds it immediately. Without an explicit selection, Add or Enter adds the ID you typed. Opening the field requests suggestions for the selected provider. Discovery does not add models to your configured list or change routing. You can still enter IDs while suggestions load, when the catalog is empty, or when listing fails.Suggestion availability
Bedrock suggestions support configured or environment bearer tokens, explicit AWS
keys, and environment AWS keys when no profile is selected. They use standard
regional endpoints. Profile-based authentication, SSO, role assumption, credential
processes, metadata credentials, and custom endpoints are not supported for
suggestions; enter model IDs manually for those configurations. These limits apply
to discovery, not to inference.
A catalog entry does not guarantee account access, tool support, or availability
through every route. Adding a model is separate from choosing its route; see
Providers for routing and credential setup.
Model fallbacks
When a model refuses to respond, Xum can retry the turn on the next model in that model’s fallback chain. Chains apply only to refusals, never to quota, auth, or network errors. Edit them under Settings → Models → Model Fallbacks, or in~/.xum/config.json under modelFallbacks.
Two cases follow Anthropic’s guidance for Claude Fable 5.2:
- Anthropic recommends no fallback for a
reasoning_extractionrefusal: the request asks the model to reproduce its reasoning in the reply. Xum shows the refusal and does not retry on the chain. Change the prompt instead. Every other refusal category, and a refusal with no category, falls back as usual. - A fallback attempt never uses Anthropic Fast mode,
even when the chat does. Anthropic limits Fast mode to some models and rejects
speed: "fast"on the others, so the fallback runs at standard speed.
New installs also get the same Opus 5 chain for the earlier Fable models,
anthropic:claude-fable-5-1 and anthropic:claude-fable-5, so an older Xum
version keeps its refusal fallback after a downgrade.
Claude Sonnet 5.5 can refuse some higher-risk cybersecurity requests, and
Anthropic’s own fallback for those requests is Claude Sonnet 5.
Updates change an existing config’s chains in one case only. When the fable
alias moves to a new Fable model, Xum adds the default chain of each Fable
model it has not seeded yet to your config, once. It never replaces a chain you already have for that model, and it
does not add the chain again after you delete it. Other new chains, such as the
Sonnet 5.5 chain, reach new installs only. If your config predates the Sonnet
5.5 chain, add it by hand:
- Under Settings → Models, add the custom model
claude-sonnet-5for Anthropic. - Under Model Fallbacks, add a chain for
anthropic:claude-sonnet-5-5and pickanthropic:claude-sonnet-5.
~/.xum/config.json directly:
One-shot Overrides
Override the model or thinking level for a single message using slash commands. The override applies only to that message — workspace settings stay unchanged.Syntax
Thinking levels
Append+level to any model alias. Levels can be named (off, low, medium/med, high, max) or numeric (0–9).
Numeric levels are model-relative — they map to the model’s allowed thinking range:
0= model’s lowest allowed level (e.g.,offfor Haiku and Sonnet,lowfor Opus or Astra)- Higher numbers select progressively higher levels, clamped to the model’s maximum
/haiku+0 disables thinking while /astra+0 sets thinking to low (Astra’s minimum).
Use /+level (no model) to override thinking on the current model: /+0 quick answer
CLI
Thexum run CLI accepts the same thinking levels via --thinking:
Next Steps
Configure Providers
Set up API keys for Anthropic, OpenAI, Google, and other providers.