Google DeepMind claims Gemini Omni 1.2 Flash cuts AI costs

Google DeepMind has announced Gemini Omni 1.2 Flash, a new AI model they say is faster and more cost-effective for handling longer audio and video clips. The company positions this release as a direct answer to business demand for processing large volumes of media without the usual wait times or high costs.

Why it matters

AI models that handle video and audio are expensive to run. Many companies have held back on using AI for longer files because of slow response times and unpredictable costs. Google DeepMind says Gemini Omni 1.2 Flash addresses both problems. If the claims hold up, this could shift the economics for businesses that rely on AI to process customer calls, security footage, or large media archives.

The announcement comes at a time when companies are looking for ways to bring more AI in-house, especially as cloud costs rise and new regulations make data residency a priority. For some, the ability to process longer clips faster could be the difference between rolling out an AI product or shelving it. This is especially relevant in sectors like media, retail, and security, where long-form audio and video are common.

What Gemini Omni 1.2 Flash actually does

According to Google DeepMind, Gemini Omni 1.2 Flash is designed to process longer clips of audio and video more quickly and at a lower cost than previous models. The company has not published detailed benchmarks or pricing, but the official blog post highlights "fast and cost-effective" as the main selling points. The model is part of the Gemini Omni line, which has targeted business users since its launch.

Images released with the announcement show a range of possible uses, from animals in conversation to a woman walking down a street, suggesting the model is intended for broad, real-world scenarios. Google DeepMind says the model can support developers building tools for media analysis, transcription, and automated content review. There is no mention of support for on-device processing, so this appears to be a cloud offering for now.

While the company is making big claims about speed and cost, there are no third-party reviews or published studies yet. Businesses considering adoption will need to test the model themselves or wait for independent validation. For those tracking the AI infrastructure market, this release follows moves by other companies to bring down the cost of running large models, as seen with Hugging Face’s Phonon TTS and OpenAI’s focus on local data centers.

What this means if you run a business

If your company deals with long audio or video files, Gemini Omni 1.2 Flash could reduce the cost and time it takes to process them. This matters for customer support teams using call analysis, media companies automating video review, or retailers analyzing in-store footage. Faster processing could also mean less lag in delivering AI-powered features to end users.

However, the lack of published benchmarks means you should approach with caution. Google DeepMind’s claims are promising, but without real-world tests, it’s hard to know how much you’ll actually save. If your business is sensitive to cloud costs or needs to keep data local, you may want to compare this offering to on-device options or providers with a local data center focus, like those discussed in OpenAI’s recent blueprint.

For now, Gemini Omni 1.2 Flash is positioned as a tool for developers and companies already invested in Google’s AI stack. If you’re considering a switch or an upgrade, watch for independent reviews and test the model against your own workloads before making a commitment.