# AI 24 Updates · 2026-11-19-2000

Published: November 19, 2026 · 19:58 CET

## In Brief

- Google Genie 4 — real-time world model now available in the Gemini API and AI Studio, with pricing
- Ideogram 4.6 — new image model with native 4K output and more precise editing
- Kling 4.0 — native 4K resolution and 10-bit HDR output
- Perplexity Agent API — xhigh effort level switches to GPT-6.2 Sol
- Microsoft AI Canvas — interactive image workspace powered by MAI Image 2.7 Flash
- Ai2 Theorizer — research-theory agent launches in Asta and goes open source

---

## 1. Google Genie 4 — real-time world model now available in the Gemini API and AI Studio, with pricing

Company: Google

Google made Genie 4 (models/genie-4), its real-time interactive world generation model, available on the paid tier of the Gemini API and in Google AI Studio.

**Facts:**
- Genie 4 seeds a world from a text prompt, a 16:9 reference image, or both, then streams steerable 24 fps 16:9 video with synchronized 48 kHz stereo audio over WebRTC and WebSockets.
- Users steer the live stream with keyboard movement controls (WASD and arrow keys) and mid-session text events, such as changing the weather or summoning characters.
- It runs on the Interactions API (v1beta) and requires a paid-tier Gemini API key; it does not support the Free tier, Batch API, Flex or Priority inference.
- Pricing is $1.50 per 1M input tokens (text or reference image) and $17.50 per 1M video output tokens (audio included), about $0.10 per second or ~$6.05 per minute of streamed output.
- Google points users to Gemini Omni 1.2 Flash for offline fixed-length clips, multi-turn video editing and upsampling to 4K.

**Sources:**
- [Gemini API docs: Genie 4](https://ai.google.dev/gemini-api/docs/models/genie-4) (official)
- [Gemini API pricing](https://ai.google.dev/gemini-api/docs/pricing) (official)
- [Google AI Studio on X](https://x.com/GoogleAIStudio/status/2123479057996116143) (official)

## 2. Ideogram 4.6 — new image model with native 4K output and more precise editing

Company: Ideogram

Ideogram released Ideogram 4.6, a new image model focused on precise editing, generating images in native 4K.

**Facts:**
- Ideogram 4.6 generates images in native 4K with no upscaling.
- Ideogram says realism and texture detail are much better, with work to remove the 'AI look' and waxy skin.
- Edits show significantly less pixel drift outside the edited area and better color and material accuracy.
- Ideogram highlights sketch-to-image, saying the model stays as faithful as possible to the input sketch.
- Ideogram also said it will skip an open 4.5 release and go directly to an open 4.6.

**Sources:**
- [Ideogram on X](https://x.com/ideogram_ai/status/2123461101224624402) (official)
- [Ideogram on X: open 4.6](https://x.com/ideogram_ai/status/2123483923632025908) (official)

## 3. Kling 4.0 — native 4K resolution and 10-bit HDR output

Company: Kling AI

Kuaishou's video model Kling 4.0 now supports native 4K resolution and 10-bit HDR output.

**Facts:**
- Kling 4.0 now generates video in native 4K resolution.
- 10-bit HDR output is available in both 4K and 1080p.
- Kling says the update brings greater detail and richer colors to every frame.
- The upgrade is available on kling.ai.

**Sources:**
- [Kling AI on X](https://x.com/Kling_ai/status/2123431119903461709) (official)

## 4. Perplexity Agent API — xhigh effort level switches to GPT-6.2 Sol

Company: Perplexity

Perplexity's Agent API xhigh effort level now runs on OpenAI's GPT-6.2 Sol with no fallback model.

**Facts:**
- The xhigh effort level now selects xhigh@2026-11-18.0, which uses openai/gpt-6.2-sol without a fallback model.
- It replaces the previous configuration of Claude Opus 5.5 with a GPT-6.1 Sol fallback.
- Tools, high reasoning effort, the 300-step limit and context compaction settings are unchanged.
- Earlier pinned versions are unchanged; pinning xhigh@2026-11-11.0 keeps the previous model configuration.

**Sources:**
- [Perplexity API changelog](https://docs.perplexity.ai/changelog/changelog) (official)

## 5. Microsoft AI Canvas — interactive image workspace powered by MAI Image 2.7 Flash

Company: Microsoft

Microsoft AI launched Canvas, an interactive image workspace on MAI Labs powered by its MAI Image 2.7 Flash model.

**Facts:**
- Canvas turns sketches into polished diagrams and supports precise edits without starting over.
- Users describe an image in a panel or start from a photo; each run makes one image in about 10 to 20 seconds.
- Parts of an image can be changed by editing a word in the prompt or selecting an object in the image.
- Canvas is available to try now on Microsoft AI's MAI Labs site.

**Sources:**
- [MAI Labs Canvas](https://labs.microsoft.ai/?model=canvas) (official)
- [Microsoft AI on X](https://x.com/MicrosoftAI/status/2123462465740533910) (official)

## 6. Ai2 Theorizer — research-theory agent launches in Asta and goes open source

Company: Ai2

Ai2 launched Theorizer, an AI agent that combines findings from scientific papers into testable theories, in its Asta research platform and open-sourced the code.

**Facts:**
- Theorizer proposes explanations for scientific observations and links each one to evidence from published studies, describing where it might apply.
- It is available in Asta, Ai2's platform of AI agents for scientific research, where Ai2 covers token costs up to a limit.
- The code is open source, so researchers can run the library locally and pay for tokens as needed.
- Scientists have tested it in their own research for nearly a year; Ai2 hosts a webinar on December 10 at 10am PT.

**Sources:**
- [Ai2 blog](https://allenai.org/blog/theorizer-in-asta) (official)
- [Ai2 on X](https://x.com/allen_ai/status/2123451767648125042) (official)
