New AI models this week: Gemini Omni 1.2 Flash and Kimi K3.1 Pro lead the roundup

Google’s Gemini Omni 1.2 Flash is the most important release of the week of November 1 to 8, 2026: a video model that now makes 15-second clips and costs 30% less than its predecessor. Moonshot AI’s open-weight Kimi K3.1 Pro was also picked up on Hugging Face, and Z.ai added a GLM 5.3 variant to OpenRouter’s catalogue. Prices below are in US dollars per 1M tokens and come from our AI API pricing tracker, which reads OpenRouter.

Gemini Omni 1.2 Flash: Google’s video model gets longer clips and a lower price

Google DeepMind announced Gemini Omni 1.2 Flash on November 5 and describes it as generally available, the first time the Gemini Omni line has reached that status. Clips can now run up to 15 seconds, up from 10 with Omni 1.1, and the length can be set from 3 to 15 seconds in one-second steps. Google lists 720p video at $0.07 per second, which it says is 30% below Omni 1.1. At that rate a 15-second 720p clip costs about $1.05. Google’s post gives only the 720p rate, so 1080p and 4K prices need checking on its pricing page.

The model accepts character, product or style images, motion clips of up to three seconds and audio cues as references. It also supports conversational edits such as swapping a product or background, changing lighting or switching aspect ratio without regenerating the clip. Google claims better lip-sync and speaker separation and fewer unwanted jump cuts in camera moves. It lists fast sports action, rapid script changes and complex physics as open problems. These claims come from Google’s own examples, with no independent benchmarks. Outputs carry a SynthID watermark.

Developers can reach the model through Google AI Studio and the Gemini API, and businesses through Google Cloud, where Provisioned Throughput is offered. It is also rolling out in Flow, Google Vids (15-second clips at no extra cost), Google Ads, and the Gemini app, starting with Google AI Ultra subscribers. OpenRouter lists it at $1.50 for input and $9.00 for output, with a 131,072-token context window. Read the official announcement, or our report on the price cut.

Kimi K3.1 Pro: Moonshot AI’s open-weight agentic model

Kimi K3.1 Pro is the first model in Moonshot AI’s K3.1 series, post-trained from Kimi K3 with large-scale reinforcement learning. Its model card calls it natively multimodal, with text and image input. It is a mixture-of-experts model with 2.8T total parameters and 104B active per token, and it has a context length of 1,048,576 tokens. Moonshot says it improves coding, 3D creation, financial research and factual accuracy.

The model card reports scores such as 94.2 on GPQA-Diamond, 92.1 on BrowseComp and 65.7 on Terminal-Bench 4.0. All of them are self-reported by Moonshot and have not been independently verified. The weights are released under the Kimi K3.1 Pro License, so the terms should be read before commercial use. The card recommends vLLM, SGLang and TokenSpeed for self-hosting, and the model is available through Moonshot’s API as kimi-k3.1-pro. Its reasoning effort can be set to low, high or max, with max as the default. The card gives no exact release date, and no price appeared in our tracker data this week. Details are on the Hugging Face model card.

GLM 5.3 FlashX Legal: A new Z.ai variant on OpenRouter

OpenRouter added z-ai/glm-5.3-flashx-legal on November 2, with Z.ai named as the provider. It costs $0.37 for input and $1.25 for output and has a context window of 1,048,576 tokens. The listing does not say what the “Legal” in the name refers to. It states no license, and it links to no official Z.ai page. No capability or benchmark claims can be made until Z.ai publishes documentation. Until then, the OpenRouter listing is the only source.

The week in numbers

• Models covered: 3 (Gemini Omni 1.2 Flash, Kimi K3.1 Pro, GLM 5.3 FlashX Legal).
• Cheapest per 1M input tokens: GLM 5.3 FlashX Legal at $0.37.
• Most expensive per 1M input tokens: Gemini Omni 1.2 Flash at $1.50. Kimi K3.1 Pro had no price in our data.
• Largest context window: GLM 5.3 FlashX Legal and Kimi K3.1 Pro, both at 1,048,576 tokens.