Greptile Summary

This automated sync adds four new models (DeepSeek V4 Pro, DeepSeek V4 Flash, Kimi K2.6, GPT-5.5), enriches existing Qwen models with detailed reasoning capability metadata, adds alias metadata for DeepSeek Reasoner/Chat, moves the Meta provider block earlier in the file, and refreshes ByteDance model timestamps.

• One U+FFFD character corruption was introduced in doubao-seed-1-6-vision's raw_limits entry — \"最大思维链: 32k\" became \"最大思\\uFFFD\\uFFFD链: 32k\" — causing garbled text in the UI for that field.

Confidence Score: 4/5

Safe to merge after fixing the single character encoding corruption in doubao-seed-1-6-vision metadata.

One P1 data-integrity issue: a U+FFFD replacement character pair corrupts a display string for one ByteDance model. All other changes (new models, capability enrichments, timestamp updates) look structurally correct. The fix is a one-line correction.

frontend/src/features/models/data/providers.json — line 13669

Important Files Changed

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[upstream all.json] -->|weekly sync workflow| B[providers.json sync]
    B --> C{Changes}
    C --> D[New models\nDeepSeek V4 Pro/Flash\nKimi K2.6\nGPT-5.5]
    C --> E[Enriched reasoning metadata\nQwen models\nsupported / interleaved / summaries / visibility]
    C --> F[Alias metadata\nDeepSeek Reasoner → deepseek-v4-flash\nDeepSeek Chat → deepseek-v4-flash]
    C --> G[Meta provider block\nrepositioned in file]
    C --> H[ByteDance models\ntimestamp + capability updates]
    H --> I[⚠️ doubao-seed-1-6-vision\nU+FFFD corruption in raw_limits]

_{Reviews (1): Last reviewed commit: "chore: sync model developers data" | Re-trigger Greptile}