October 21, 2026
October 2026 model refresh: what is new and what changed
New models for pictures, video, voice, 3D and chat arrive on 21 October. Some older models, including Veo 3.1, are replaced by newer ones.
This is our October 2026 model refresh. A model is the AI that makes your picture, video or voice. We added new ones, replaced some older ones, and fixed a few that were failing.
What the model refresh adds
These new models are in the model list today:
- Chat and writing: Gemini 4 Argon and Gemini 3.8 Flash from Google, GPT-6.1 Sol and GPT-6 Luna from OpenAI, and DeepSeek Flash.
- Pictures: Gemini Nano Banana 2.1, FLUX 3, Krea 3, Ideogram 4.5, Seedream 5 Flash and Recraft 4.1 Flash.
- Voice: two new Gemini 3.8 voices that read text aloud.
- 3D: Meshy 7.1 and Tripo P2. Both can make a 3D model from words or from a picture.
- Video: Luma Ray 3.2, Kling v3 Standard, Kling O3 Pro Reference, FLUX 3 Max video and Vidu Q4.
- Talking portraits: MiniMax H3 Max and OmniHuman 2 make a still portrait speak.
- Sound effects: Mirelo SFX 2.0.
Each new model is priced from what its maker charges. Talking portraits need a voice clip of the right length. MiniMax H3 Max needs 5 to 10 seconds, and OmniHuman 2 needs 4 to 10 seconds. If your clip is a different length, genvi stops the run before it starts and gives your credits back.
Older models that changed
- Veo 3.1 (video). Google shuts down three Veo 3.1 versions on 22 October. Anything you saved with Veo now uses Grok Imagine Video 1.5 Lite instead, and you pay the Veo price you picked. Lite makes 720p video, so a 1080p or 4K request comes out at 720p.
- Gemini 2.5 Flash Image (pictures). Now runs on Nano Banana 2.1.
- Luma Ray 2 Flash (video). Replaced by Ray 3.2, because the company fal shuts it down on 25 October.
- FLUX.1 [pro] (pictures). Now runs on FLUX.2 [pro].
- Going away soon. Gemini 3.1 Flash Image, Lyria 3 Pro and two preview voices are on their way out. Google ends the voices on 17 November. The TTS-1 and TTS-1 HD voices end on 6 January 2027.
Speech steps and the agent's voice-overs now start with Gemini 3.8 Flash Lite TTS, one of the new voices.
A vintage sports car on a mountain bend, made with GPT Image 2.5 Flare.
Repairs
Some video and 3D models were failing after we had already held your credits, because the connection to them was broken. We fixed LTX 2.3 quality, two Kling 2.5 Standard models, and image-to-video on Kling 2.6, v3 and O3. The Meshy Texture Generator now uses Meshy 7 Retexture.
Not available yet
Google says Gemini Omni Flash is the replacement for Veo. We have built support for it but have not tested it with a real run yet, so it is not in the list today. We will tell you when it is ready.
What it costs
You see the price in the model list and in the estimate before you run. Credits follow the maker's price, rounded up.