The Short Answer
Ghotly AI 4.3.2, released on 19 November 2026, is a free update for the Windows app. We checked every AI provider's model list on the same day and updated the app to match: new models from Google, OpenAI and Anthropic, a new NVIDIA list (NVIDIA retired the old one on 15 and 16 November), and a refreshed list of OpenRouter's free models. When a model is slow or busy, the app now asks that provider's quicker model instead of making you wait.
To get it, open the app, let the update download, then close the app so it installs. Or download the installer from the Download page.
New AI Models
You still use your own API keys. The model you chose stays chosen, unless its provider has retired it: then the app uses that provider's first model in the list.
| Provider | What changed |
|---|---|
| Google Gemini | New: Gemini 4 Hydrogen (Google's newest, free tier) and Gemini 3.5 Flash-Lite (free tier). Still there: 3.6 Flash, 3.8 Flash, 3.1 Flash-Lite |
| OpenAI | New: GPT-6.2 Sol (now first in the list) and GPT-6.2 Astra. GPT-4o is gone from the list: OpenAI is shutting it down |
| Anthropic | New: Claude Fable 5.2, Anthropic's most capable model (slower and more expensive) |
| NVIDIA | A new list from NVIDIA's live catalog: Nemotron 3.5 Lightning, Mistral Large 4, GLM 5.3 Flash, MiniMax M3.1 Flash, Kimi K3.1 Pro and more |
| OpenRouter | The free list refreshed: 14 free models, and Auto now picks only free models |
| Groq, Grok | No change: GPT-OSS 120B and 20B on Groq, Grok 4.7 to 4.3 on xAI |
Fixes
- NVIDIA works again. NVIDIA retired all three models the app used on 15 and 16 November 2026, so every NVIDIA answer failed. The new models are in, and when NVIDIA retires one again, the app tries the next one in the list by itself.
- No more "requires more credits" on a free OpenRouter key. OpenRouter's Auto could pick paid models. Auto now uses OpenRouter's router for free models only, and a screenshot always goes to a free model that can read it.
- Answers that waited a minute. At busy hours, Gemini 3.6 Flash sometimes sent nothing for a whole minute. Now, when a model has not started answering after 15 seconds, the app asks the same provider's quicker model with the same key, for example Gemini 3.1 Flash-Lite. The screen tells you when this happens.
- A busy model. When a provider says a model is busy, the app asks its quicker model after two seconds, instead of the same busy one.
- GPT-6 answers start sooner. GPT-6 models now get the same "low" reasoning setting as GPT-5, so the first words come sooner.
- Screenshots with only a Groq key. The app now tries Groq's preview model that can read images. A free Gemini key is still the best choice for Screen Analysis.
A New Shortcut
Ctrl + Shift + Up / Down scrolls a long answer without clicking into the window. Hold the keys to keep scrolling. Like every shortcut, you can change it in Settings > Shortcuts. See all of them on the Shortcuts page.
Common Questions
Is Ghotly AI 4.3.2 free?
Yes. The update is free for every account, and the app downloads it by itself. Your plan, your minutes and your saved keys stay as they are.
Do I have to choose a new model after the update?
Only if you want to. Your chosen model stays, unless its provider has retired it, like NVIDIA's old models or GPT-4o. Then the app uses that provider's first model, and you can pick another in Settings.
Which model should I use for interviews?
For speed, Gemini 3.6 Flash or Gemini 3.1 Flash-Lite with a free Gemini key, or GPT-OSS 20B on Groq. For the hardest questions, Gemini 4 Hydrogen, GPT-6.2 Astra or Claude Fable 5.2: they take longer to start and the paid ones cost more per answer. Our model guide has the measured speeds and the cost of one answer.
Why did the app say it asked a quicker model?
The model you chose had not started answering after 15 seconds, or its provider said it was busy. Rather than leave you waiting in the middle of an interview, the app asked the same provider's quicker model with the same key.
