[rdself]

Summary

• Make xhigh reasoning effort pass through by default unless a model explicitly sets supportsXHighEffort: false.
• Keep the existing max normalization behavior separate from the xhigh pass-through policy.
• Update sanitize and provider model tests for OpenRouter DeepSeek, explicit Claude opt-outs, and existing max behavior.

Validation

• node --import tsx/esm --test tests/unit/provider-models-config.test.ts tests/unit/base-executor-sanitize-effort.test.ts tests/unit/executor-base-utils.test.ts
• npm run typecheck:core
• git diff --check

[Copilot]

Pull request overview

Note

Copilot was unable to run its full agentic suite in this review.

This PR updates “xhigh” reasoning-effort handling to be opt-out (pass-through by default), while preserving the older opt-in behavior specifically for max → xhigh normalization.

Changes:

• Switch supportsXHighEffort to treat supportsXHighEffort: false as the only opt-out signal (otherwise pass-through).
• Add supportsXHighEffortForMaxNormalization and update the sanitizer to use it when deciding whether to normalize max to xhigh.
• Update unit tests to reflect the new pass-through-by-default behavior for xhigh.

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
tests/unit/provider-models-config.test.ts	Adds coverage for new max-normalization gate and broader xhigh pass-through expectations.
tests/unit/executor-base-utils.test.ts	Updates expectation that xhigh is preserved unless explicitly unsupported.
tests/unit/base-executor-sanitize-effort.test.ts	Expands/updates sanitizer tests to match opt-out xhigh behavior and explicit downgrade cases.
open-sse/executors/base.ts	Uses new max-normalization gate and updates sanitizer doc comment.
open-sse/config/providerModels.ts	Implements opt-out xhigh logic and introduces new helper for max normalization.

[gemini-code-assist]

Code Review

This pull request changes the reasoning effort sanitization logic so that xhigh effort is allowed by default (pass-through) across providers unless a model explicitly opts out, while preserving the original opt-in behavior for max to xhigh normalization via a new helper supportsXHighEffortForMaxNormalization. Feedback on the changes highlights that dynamic or custom Anthropic-compatible providers (starting with anthropic-compatible-) will bypass safety checks because they lack a static mapping to the cc alias, and suggests resolving these prefixes to cc to ensure correct capability lookups.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

[gemini-code-assist]

For dynamic or custom Anthropic-compatible providers (whose IDs start with anthropic-compatible-), PROVIDER_ID_TO_ALIAS will not contain a static mapping, causing supportsXHighEffort and supportsXHighEffortForMaxNormalization to treat them as unknown providers. This defaults them to true (pass-through) for all models, bypassing safety checks and allowing unsupported effort levels (like xhigh or max) on models that explicitly opt out (e.g., claude-haiku-4-5-20251001).

Resolving any anthropic-compatible- prefix to "cc" (the alias for "claude") ensures we correctly look up the model's capabilities in the Claude registry.

  // Keep explicit false entries as the unsupported-model list. Unlisted models
  // and models without an explicit flag pass through unchanged.
  const resolvedId = aliasOrId.startsWith("anthropic-compatible-") ? "cc" : aliasOrId;
  const resolvedAlias = PROVIDER_ID_TO_ALIAS[resolvedId] || resolvedId;
  const resolvedModel = getProviderModel(resolvedAlias, modelId);
  return resolvedModel?.supportsXHighEffort !== false;
}

export function supportsXHighEffortForMaxNormalization(
  aliasOrId: string,
  modelId: string
): boolean {
  const resolvedId = aliasOrId.startsWith("anthropic-compatible-") ? "cc" : aliasOrId;
  const alias = PROVIDER_ID_TO_ALIAS[resolvedId] || resolvedId;
  if (alias === "cc") {
    return supportsXHighEffort(aliasOrId, modelId);
  }
  const providerModels = PROVIDER_MODELS[alias] || PROVIDER_MODELS[resolvedId];
  if (!providerModels) return true;
  const model = getProviderModel(alias, modelId);
  return model?.supportsXHighEffort === true;
}

[diegosouzapw]

Thanks @rdself! Splitting the xhigh policy into opt-out pass-through (supportsXHighEffort) while keeping the conservative opt-in behavior isolated for max-normalization (supportsXHighEffortForMaxNormalization) is a clean separation, and the tests are stronger for it (the previous xhigh || high OR-assertions were masking). Validated 49/49 on the merged state + typecheck:core clean. Merging into release/v3.8.24 for the upcoming release. 🙌