feat: add GPT-5.4 support and auto-fallback for future GPT-5.x sub-versions by EurFelux · Pull Request #13293 · CherryHQ/cherry-studio

EurFelux · 2026-03-07T18:01:02Z

What this PR does

Before this PR:

isGPT5SeriesModel used a manual exclusion list (!includes('gpt-5.1') && !includes('gpt-5.2')) that required updating for every new sub-version.
GPT-5.2 had a dedicated branch in _getThinkModelType identical to the fallback logic.
Future GPT-5.x models (e.g., gpt-5.4) would fall through to the default thinking type with no reasoning effort support.
GPT-5.2 codex models incorrectly inherited none reasoning effort from the gpt5_2 type.
Verbosity support excluded all chat/codex variants uniformly, without version-aware logic.
gpt-5-pro verbosity was restricted to high only, which no longer reflects the latest API behavior.

After this PR:

isGPT5SeriesModel uses a regex negative lookahead (gpt-5(?!\.\d)) to automatically exclude all sub-versions.
Added isGPT5FamilyModel to match the entire GPT-5 family (base + all sub-versions).
Refactored isSupportVerbosityModel, isSupportNoneReasoningEffortModel, and isSupportedReasoningEffortOpenAIModel to use isGPT5FamilyModel.
_getThinkModelType uses isGPT5FamilyModel as a parent guard with an automatic fallback for unknown sub-versions (routes to gpt5_2/gpt52pro).
Added gpt5_2_codex thinking model type with [low, medium, high, xhigh] (no none).
GPT-5.3+ codex models support none reasoning effort via isSupportNoneReasoningEffortModel.
Verbosity logic is now version-aware: chat/codex variants before 5.3 fall back to medium only, while 5.3+ codex supports all levels.
gpt-5-pro verbosity updated from high only to [low, medium, high], matching the latest API test results.
Added comprehensive test coverage for all changes.

Why we need it and why it was done in this way

The following tradeoffs were made:

Future GPT-5.x sub-versions (5.3+) fall back to GPT-5.2 reasoning configuration, which may not be perfectly accurate but provides reasonable defaults until explicit support is added.

The following alternatives were considered:

Adding explicit isGPT54SeriesModel checks (rejected: doesn't scale, same maintenance burden as before).
Using a version parsing approach (rejected: over-engineering for the current use case).

Breaking changes

None. All existing model detection behavior is preserved. Only new models (GPT-5.3+) gain automatic fallback support.

Behavior change: gpt-5-pro verbosity expanded from high only to [low, medium, high]. This reflects updated API capabilities confirmed by testing — previously the API only supported high, but now supports additional levels.

Special notes for your reviewer

The isGPT5FamilyModel function is intentionally simple (modelId.includes('gpt-5')) — it serves as a broad family guard, with specific sub-version checks handled inside.
GPT-5.2 codex gets its own explicit branch and type (gpt5_2_codex) because it supports xhigh but not none, which doesn't match any existing codex type.
isSupportVerbosityModel is now simply isGPT5FamilyModel — granular exclusions (chat/codex) are handled by MODEL_SUPPORTED_VERBOSITY validators with a fallback to medium.
gpt-5-pro verbosity change is intentional — latest API testing confirms it now supports low, medium, and high (previously only high).

Checklist

PR: The PR description is expressive enough and will help future contributors
Code: Write code that humans can understand and Keep it simple
Refactor: You have left the code cleaner than you found it (Boy Scout Rule)
Upgrade: Impact of this change on upgrade flows was considered and addressed if required
Documentation: A user-guide update was considered and is present (link) or not required. Check this only when the PR introduces or changes a user-facing feature or behavior.
Self-review: I have reviewed my own code (e.g., via /gh-pr-review, gh pr diff, or GitHub UI) before requesting review from others

Release note

Add GPT-5.4 model support with automatic reasoning effort configuration. Future GPT-5.x sub-versions will also be automatically supported with sensible defaults.

Replace manual exclusion list with regex `gpt-5(?!\.\d)` to automatically exclude all sub-versions (gpt-5.1, gpt-5.2, etc.). Co-Authored-By: Claude Opus 4.6 <[email protected]> Signed-off-by: icarus <[email protected]>

Add isGPT5FamilyModel to match all GPT-5 family models (gpt-5, gpt-5.1, gpt-5.2, etc.). Expand test coverage for both isGPT5FamilyModel and isGPT5SeriesModel with future sub-version cases. Co-Authored-By: Claude Opus 4.6 <[email protected]> Signed-off-by: icarus <[email protected]>

Signed-off-by: icarus <[email protected]>

Cover codex exclusion and future GPT-5.x sub-version scenarios for isSupportVerbosityModel, isSupportNoneReasoningEffortModel, and isSupportedReasoningEffortOpenAIModel. Co-Authored-By: Claude Opus 4.6 <[email protected]> Signed-off-by: icarus <[email protected]>

Simplify isSupportVerbosityModel, isSupportNoneReasoningEffortModel, and isSupportedReasoningEffortOpenAIModel by replacing manual isGPT5Series || isGPT51 || isGPT52 checks with isGPT5FamilyModel. Add codex exclusion for verbosity and none-reasoning-effort models. Co-Authored-By: Claude Opus 4.6 <[email protected]> Signed-off-by: icarus <[email protected]>

fallback

Allow codex variants from GPT-5.3 onwards to use reasoning_effort "none". GPT-5.1/5.2 codex models remain excluded. Co-Authored-By: Claude Opus 4.6 <[email protected]> Signed-off-by: icarus <[email protected]>

GPT-5.2 codex does not support none reasoning effort. Add a dedicated gpt5_2_codex type with [low, medium, high, xhigh] options, while GPT-5.3+ codex falls back to gpt5_2 which includes none. Co-Authored-By: Claude Opus 4.6 <[email protected]> Signed-off-by: icarus <[email protected]>

Update tests to match isSupportVerbosityModel now being isGPT5FamilyModel. Chat/codex exclusion is handled by MODEL_SUPPORTED_VERBOSITY validators: - Chat models get medium only - Old codex (5.1/5.2) gets medium only - New codex (5.3+) gets all levels - Pro models get all levels Co-Authored-By: Claude Opus 4.6 <[email protected]> Signed-off-by: icarus <[email protected]>

EurFelux · 2026-03-07T18:49:16Z

NOTE: text verbosity and reasoning effort have been tested across all variants of the GPT 5 family models via OpenAI's official responses API.

Replace `() => true` with `isGPT5FamilyModel` for clearer semantics. Co-Authored-By: Claude Opus 4.6 <[email protected]> Signed-off-by: icarus <[email protected]>

EurFelux

Self-Review Summary 🐱

Overall: Looks good! The architecture is clean and the approach is well-thought-out. CI passes (basic-checks ✅, render-test ✅, skills-check-windows ✅).

What I like

Regex negative lookahead for isGPT5SeriesModel is elegant and future-proof — no more manual exclusion lists
Two-tier architecture (isGPT5FamilyModel as broad guard + sub-version checks) makes fallback behavior natural
Decoupling isSupportVerbosityModel from granular level control is a good separation of concerns
Comprehensive test coverage for all the new code paths, including future model scenarios

Minor items (all nits)

Dead first validator in MODEL_SUPPORTED_VERBOSITY — the !isSupportVerbosityModel check is already handled by the early return before the loop
Outdated JSDoc on getModelSupportedVerbosity — still references old behavior (gpt-5-pro = high only)
Ambiguous comments — "5.2 and after 5.x" and "Fallback for GPT-5.x sub-versions (5.2+)" could be clearer about what they cover

No blocking issues found. Ready for external review.

src/renderer/src/config/models/utils.ts

src/renderer/src/config/models/reasoning.ts

DeJeune

Review Summary

Well-structured PR that improves GPT-5.x model detection scalability. The regex negative lookahead for isGPT5SeriesModel and the isGPT5FamilyModel umbrella guard are good design choices. Test coverage is comprehensive and the PR description is excellent.

Critical / Significant

isSupportNoneReasoningEffortModel vs MODEL_SUPPORTED_REASONING_EFFORT inconsistency — After this PR, isSupportNoneReasoningEffortModel returns false for gpt-5.1-codex and gpt-5.1-codex-max, but their reasoning effort options in reasoning.ts still include none. Users can select none in the UI, but the parameter won't be sent to the API (falls through to a warning log in aiCore/utils/reasoning.ts:157). Either remove none from the options or adjust the isOldCodex guard to only exclude GPT-5.2 codex.
isSupportVerbosityModel behavioral change impacts multiple consumers — Now returns true for chat/codex models (previously false). This affects OpenAIResponseAPIClient.ts (now sends text.verbosity for chat models), and UI components (shows a verbosity dropdown with only medium for chat models). Please verify the API tolerates the text.verbosity field for chat endpoints.

Minor / Nits

isGPT5FamilyModel could false-match hypothetical model names like gpt-50.
Missing test for base gpt-5.2 verbosity (non-codex, non-pro, non-chat).

Positives

Regex negative lookahead for isGPT5SeriesModel is a clean, scalable solution.
isGPT5FamilyModel + specific sub-version checks inside _getThinkModelType is a good architectural pattern.
Automatic fallback for GPT-5.3+ to gpt5_2/gpt52pro types is well thought out.
Thorough test coverage for all new edge cases and future model variants.
Excellent PR description with clear rationale and tradeoffs.

src/renderer/src/config/models/openai.ts

src/renderer/src/config/models/reasoning.ts

src/renderer/src/config/models/openai.ts

src/renderer/src/config/models/utils.ts

Improve comment clarity for fallback logic and update outdated JSDoc for getModelSupportedVerbosity to reflect version-aware behavior. Co-Authored-By: Claude Opus 4.6 <[email protected]> Signed-off-by: icarus <[email protected]>

GPT-5.1 codex and codex-max do not support reasoning_effort: none. Remove none from their option lists and update tests accordingly. Also add missing gpt-5.2 base verbosity test case. Co-Authored-By: Claude Opus 4.6 <[email protected]> Signed-off-by: icarus <[email protected]>

GeorgeDong32

Code Review Summary

Overall: APPROVED ✅

Excellent PR with a clean architecture and comprehensive test coverage. The regex negative lookahead for isGPT5SeriesModel is elegant and future-proof.

Verified

✅ Lint checks pass
✅ All 2521 tests pass
✅ TypeScript compilation successful
✅ Issue #2 (verbosity API compatibility) - Confirmed by EurFelux that OpenAI API was tested

Minor Suggestions (Optional)

Dead code in MODEL_SUPPORTED_VERBOSITY (utils.ts:225-229)
- The first validator !isSupportVerbosityModel is never reached since there's already an early return in getModelSupportedVerbosity
- Consider removing it for cleaner code
Potential false match in isGPT5FamilyModel (openai.ts:70-73)
- Current: modelId.includes('gpt-5') could match hypothetical models like gpt-50, gpt-500
- Suggestion: Consider using a more precise regex like /^gpt-5(\.\d+)?(-[\w-]+)?$/i
Missing test case
- Base gpt-5.2 verbosity test (non-codex, non-pro, non-chat) is covered in the diff but could be explicitly added

Positives

Regex negative lookahead is clean and scalable
Two-tier architecture (isGPT5FamilyModel + sub-version checks) is well-designed
Automatic fallback for GPT-5.3+ is a smart approach
Excellent PR description with clear rationale

Great work! Ready to merge. 🚀

…overage Co-Authored-By: Claude Opus 4.6 <[email protected]> Signed-off-by: icarus <[email protected]>

EurFelux · 2026-03-08T06:41:36Z

Thanks for the review @GeorgeDong32! 🙏

Regarding your suggestions:

Dead code in MODEL_SUPPORTED_VERBOSITY — Keeping it as a defensive guard. The UI is gated by isSupportVerbosityModel anyway, so even if triggered, it wouldn't surface to users.
isGPT5FamilyModel false match — Given current OpenAI naming conventions, gpt-50 is extremely unlikely. Keeping the simple includes for readability — we can tighten it if needed in the future.
Missing test case — Added! Reorganized verbosity tests by model version (5.0 → 5.4) with full variant coverage (base, pro, chat, codex), plus a future fallback test for GPT-5.5+. See 7e61caf.

### What this PR does Before this PR: - Version is 1.7.24 After this PR: - Version is 1.7.25 - Release notes updated in `electron-builder.yml` ### Why we need it and why it was done in this way Standard release workflow: collect commits since v1.7.24, generate bilingual release notes, bump version. The following tradeoffs were made: N/A The following alternatives were considered: N/A ### Breaking changes **Action Required**: The built-in filesystem MCP server now requires manual approval for write/edit/delete operations by default. ### Special notes for your reviewer **Included commits since v1.7.24:** ✨ New Features: - feat(websearch): add Querit search provider (#13050) - feat(minapp,provider): add MiniMax Agent, IMA mini apps and MiniMax Global, Z.ai providers (#13099) - feat(provider): add agent support filter for provider list (#11932) - feat(agent): add custom environment variables to agent configuration (#13357) - feat: add GPT-5.4 support (#13293) - feat(gemini): add thought signature persistence (#13100) 🐛 Bug Fixes: - fix: secure built-in filesystem MCP root handling (#13294) - fix: check underlying tool permissions for hub invoke/exec (#13282) - fix: remove approval countdown timers and add system notifications (#13281) - fix: agent tool status not stopping on abort (#13111) - fix: resolve spawn ENOENT on Windows for Code Tools (#13405) - fix: Use default assistant for topic auto-renaming (#13387) - fix(backup): defer auto backup during streaming response (#13307) - fix(ui): fix Move To submenu overflow (#13399) - fix: render xml fenced svg blocks as svg previews (#13431) - fix: correct Gemini reasoning params (#13388) - fix: improve Qwen 3.5 reasoning model detection (#13235) - fix: upgrade xAI web search to Responses API (#12812) - fix(api-server): relax chat completion validation (#13279) - fix: install or uninstall button state issues (#13114) - fix: improve update dialog behavior (#13363) - fix: auto-convert reasoning_effort to reasoningEffort (#12831) ### Checklist - [x] PR: The PR description is expressive enough and will help future contributors - [x] Code: Write code that humans can understand and Keep it simple - [x] Refactor: You have left the code cleaner than you found it (Boy Scout Rule) - [x] Upgrade: Impact of this change on upgrade flows was considered and addressed if required - [x] Documentation: A user-guide update was considered and is present (link) or not required. - [x] Self-review: I have reviewed my own code before requesting review from others ### Release note ```release-note See release notes in electron-builder.yml ``` --------- Signed-off-by: kangfenmao <[email protected]>

EurFelux and others added 10 commits March 7, 2026 23:35

refactor: use regex negative lookahead for isGPT5SeriesModel

e58fc0a

Replace manual exclusion list with regex `gpt-5(?!\.\d)` to automatically exclude all sub-versions (gpt-5.1, gpt-5.2, etc.). Co-Authored-By: Claude Opus 4.6 <[email protected]> Signed-off-by: icarus <[email protected]>

fix: Update GPT-5 verbosity configuration logic

39c4016

Signed-off-by: icarus <[email protected]>

feat: Add GPT-5.2 reasoning model support and future sub-version

c00e0ac

fallback

feat: support none reasoning effort for GPT-5.3+ codex models

15c371a

Allow codex variants from GPT-5.3 onwards to use reasoning_effort "none". GPT-5.1/5.2 codex models remain excluded. Co-Authored-By: Claude Opus 4.6 <[email protected]> Signed-off-by: icarus <[email protected]>

fix: Correct OpenAI verbosity model support logic

bfa2d3c

EurFelux and others added 2 commits March 8, 2026 02:51

refactor: use isGPT5FamilyModel for verbosity fallback validator

5134a23

Replace `() => true` with `isGPT5FamilyModel` for clearer semantics. Co-Authored-By: Claude Opus 4.6 <[email protected]> Signed-off-by: icarus <[email protected]>

fix: Filter out non-verbosity models from verbosity validation

b36e9bb

EurFelux commented Mar 7, 2026

View reviewed changes

src/renderer/src/config/models/utils.ts Show resolved Hide resolved

src/renderer/src/config/models/utils.ts Show resolved Hide resolved

src/renderer/src/config/models/reasoning.ts Outdated Show resolved Hide resolved

src/renderer/src/config/models/reasoning.ts Outdated Show resolved Hide resolved

DeJeune requested changes Mar 7, 2026

View reviewed changes

EurFelux and others added 2 commits March 8, 2026 03:10

docs: clarify GPT-5.x reasoning and verbosity comments

33b7154

Improve comment clarity for fallback logic and update outdated JSDoc for getModelSupportedVerbosity to reflect version-aware behavior. Co-Authored-By: Claude Opus 4.6 <[email protected]> Signed-off-by: icarus <[email protected]>

EurFelux requested a review from DeJeune March 7, 2026 19:14

DeJeune approved these changes Mar 7, 2026

View reviewed changes

feat: Update OpenAI model list with latest GPT versions

fa11af0

GeorgeDong32 reviewed Mar 8, 2026

View reviewed changes

test: reorganize verbosity tests by model version with full variant c…

7e61caf

…overage Co-Authored-By: Claude Opus 4.6 <[email protected]> Signed-off-by: icarus <[email protected]>

EurFelux requested a review from GeorgeDong32 March 8, 2026 06:42

EurFelux linked an issue Mar 8, 2026 that may be closed by this pull request

Cherry does not support GPT-5.4 (xhigh) in version 5.4 #13287

Closed

GeorgeDong32 approved these changes Mar 8, 2026

View reviewed changes

EurFelux merged commit 699682e into main Mar 8, 2026
9 checks passed

EurFelux deleted the feat/gpt-5.4 branch March 8, 2026 08:33

SiinXu mentioned this pull request Mar 11, 2026

[Bug]: GitHub Copilot model gpt-5.4 fails in Cherry Studio with: not accessible via the /chat/completions endpoint #13254

Closed

5 tasks

kangfenmao mentioned this pull request Mar 13, 2026

chore: release v1.7.25 #13438

Merged

6 tasks

SiinXu mentioned this pull request Mar 19, 2026

[Bug]: GitHub Copilot GPT-5.4 routes to /chat/completions instead of /responses endpoint #13637

Closed

10 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat: add GPT-5.4 support and auto-fallback for future GPT-5.x sub-versions#13293

feat: add GPT-5.4 support and auto-fallback for future GPT-5.x sub-versions#13293
EurFelux merged 16 commits intomainfrom
feat/gpt-5.4

EurFelux commented Mar 7, 2026 •

edited

Loading

Uh oh!

EurFelux commented Mar 7, 2026

Uh oh!

EurFelux left a comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

DeJeune left a comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

GeorgeDong32 left a comment

Uh oh!

EurFelux commented Mar 8, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Conversation

EurFelux commented Mar 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What this PR does

Why we need it and why it was done in this way

Breaking changes

Special notes for your reviewer

Checklist

Release note

Uh oh!

EurFelux commented Mar 7, 2026

Uh oh!

EurFelux left a comment

Choose a reason for hiding this comment

Self-Review Summary 🐱

What I like

Minor items (all nits)

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

DeJeune left a comment

Choose a reason for hiding this comment

Review Summary

Critical / Significant

Minor / Nits

Positives

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

GeorgeDong32 left a comment

Choose a reason for hiding this comment

Code Review Summary

Verified

Minor Suggestions (Optional)

Positives

Uh oh!

EurFelux commented Mar 8, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

EurFelux commented Mar 7, 2026 •

edited

Loading