Newest models and launch day support in Zoo Code v3.78
🦓 Zoo Code v3.78 is here!
This release brings three major new models, NanoGPT provider support, and stronger provider and task reliability.
✨ What’s new
Three major new models have arrived: Gemini 3.7 Flash, Qwen3.8 Max, and GLM 5.3 with launch day support, made possible through new collaborations with leading AI labs.
DeepSeek V4 updates: Improved reasoning mappings, corrected pricing, and expanded provider coverage for the V4 Pro 0813 checkpoint.
NanoGPT provider support: Connect with dynamic model discovery, streaming and prompt completions, and routing preferences for speed, price, latency, throughput, tool support, and caching.
More reliable Azure OpenAI setup: Resource endpoints are normalized correctly, with clearer Azure-specific guidance in settings.
Safer task history: Task titles are preserved when navigating rapidly before messages finish loading.
Correct Kimi Code limits: Output-token defaults and server-provided model limits are now handled correctly.
Cleaner settings migration: Zoo now uses zoo-code-settings.json as the default import/export filename.
Wider Destructive Command Guard support: DCG now works on Intel-based Macs.
Plus security updates for undici and Mermaid, telemetry improvements, provider-identifier architecture work, and a broad shared test-utilities rollout.
Full release notes:
https://github.com/Zoo-Code-Org/Zoo-Code/releases/tag/v3.78.0
Update from the VS Code Marketplace or Open VSX, and let us know what you think!