AI models move quickly. A model that is the best balance of quality, speed, and cost today may not hold that position for long.
Quendian should be able to keep up without asking users to relearn the product every few weeks. The latest update refreshes the models behind Quendian Cloud, improves the choices available to bring-your-own-key users, and gives us a safer way to keep those choices current.
New defaults, familiar workflow
The visible workflow has not changed. Start a recording, stop when the conversation is over, then receive a transcript and structured summary.
Behind that workflow, Quendian Cloud now uses upgraded transcription and summarization models with built-in fallbacks. The underlying providers and models can change as we evaluate improvements, while the experience inside Quendian remains consistent.
That combination gives us a stronger default stack while retaining a recovery path when a provider or model is unavailable.
Cloud models can improve without waiting for an app release
The most important change is architectural.
The Quendian app talks to a stable Quendian Cloud endpoint. The cloud service handles authentication, privacy controls, quotas, usage metering, model selection, and fallback behavior on the server.
This means we can evaluate and improve the Quendian Cloud summarization route without requiring everyone to download a new version of the Mac app. We evaluate those changes using our own test data, never your recordings, transcripts, summaries, or notes. The cloud architecture also gives us one place to observe failures, verify usage, and roll back a model change if it does not perform as expected.
Faster upgrades should not mean impulsive upgrades. Before a new default is promoted, we can test the complete path: streaming output, fallback behavior, usage accounting, error handling, and rollback. A newer model still has to work reliably inside the product.
Better choices for bring-your-own-key users
Quendian’s bring-your-own-key mode remains an important part of the product. Your provider credential stays on your Mac, and requests go directly to the provider you configure.
The latest app update adds provider-specific model choices for OpenAI, Anthropic, Groq, and Together. The recommended defaults now reflect current model families, including GPT-5.6 Terra for OpenAI, Claude Sonnet 5 for Anthropic, Llama 3.3 70B Versatile for Groq, and GLM-5.1 for Together.
Those menus are guidance, not a new lock-in. You can still enter a custom model ID, use OpenRouter, or connect compatible local services such as Ollama and LM Studio. Existing custom and older model selections are preserved rather than silently replaced.
Quendian also checks whether the configured key is accepted and, when the provider exposes a standard model list, whether the selected model is available to that account. That turns a confusing failure during summarization into a clear configuration issue while you are still in Preferences.
The privacy boundary stays the same
Changing a model should not quietly change who receives your data.
In Quendian Cloud mode, processing continues through Quendian’s documented cloud infrastructure and configured providers. In bring-your-own-key mode, Quendian continues to send the request directly from the Mac to the provider selected by the user. Local model options remain local.
The upgrade changes model selection and routing, not the ownership model behind the product.
Keeping pace without chasing every release
There will always be another model announcement. The goal is not to add every new name to a menu on launch day.
The goal is to keep Quendian’s defaults current, evaluate meaningful improvements quickly, preserve fallbacks, and leave room for users who want a different provider or model. Quendian Cloud should improve quietly behind a stable interface, while bring-your-own-key users should retain explicit control.
Models move quickly. Quendian’s upgrade path now can too.