Chinese AI developer SenseTime has unveiled its improved multimodal model SenseNova 5.5, claiming it represents the newest state-of-the-art.
The improved model comes just months after the discharge of SenseNova 5, which SenseTime says is on par with GPT-4 Turbo.
The improved SenseNova 5.5 with 600B parameters is claimed to represent a 30 percent improvement in overall performance.
Benchmark results released by the corporate show that its model beats Anthropic's GPT-4o and Claude Sonnet 3.5 models.
The benchmarks where SenseNova 5.5 excels are those normally used for Chinese models. If they used GPQA, Humaneval or Math benchmarks we could make a fairer comparison, besides these numbers look impressive.
SenseTime also introduced SenseNova 5o, China's first real-time multimodal model that may process text, images, audio and video.
When the SenseNova 5o was demoed on stage, it was shown that the performance may be very much like the GPT 4o demo, which we’re still waiting to check.
According to SenseTime, SenseNova 5o’s interactions are “on par with GPT-4o’s streaming interaction capabilities.”
Claude SenseNova 5.5, an AI generation with greater than 3.5/GPT-4o, was announced.
It also seems that a multimodal model, SenseNova 5o, was announced at the identical time.Performance has improved by 30% in comparison with SenseNova 5.0 and lots of core metrics, mainly math and English, exceed GPT-4o standards.pic.twitter.com/H1u98SFVwX
— Keiichi Eto|Radineer (@RadineerE10) July 8, 2024
The company also introduced a “Lite” version of SenseNova 5.5, a cloud-to-edge low-cost model that may run on the device.
SenseTime says its edge-side model will cost just RMB 9.90 per 12 months per device, but didn’t provide performance details.
As a part of SenseNova 5.5, SenseTime also released Vimi, a controllable AI avatar video generator.
Vimi can create videos up to at least one minute long using a single photo as a template and allows precise control of an avatar's facial expressions and upper body movements.
over and beyond
·Audio
·Text
·Picture
·Movie
SenseNova 5o, a multimodal real-time model that may processpic.twitter.com/CKs0JyaH1m— Aruru ChatGPT × AI Tool (@chatgptair) July 9, 2024
The Chinese exit from OpenAI
In line with U.S. sanctions on technology exports to China, OpenAI will block API access to its tools and services for users in China.
The Chinese government already blocks ChatGPT, but users have been in a position to bypass the federal government's firewall using VPNs. OpenAI has not yet provided a full explanation for this, but will block this bypass starting today.
This has led to a fierce scramble as Chinese firms search for alternatives to OpenAI's models. SenseTime announced the launch of its “Project $0 Go” program to draw users to its platform.
The program is a free and comprehensive onboarding package to assist latest enterprise users migrate from OpenAI platforms to SenseTime. It includes 50 million tokens of credit and API migration consulting services.
Other Chinese model providers are also attempting to make the most of OpenAI's exit. Baidu, Zhipu and Tencent have all offered between 50 and 150 million tokens as an incentive to migrate to their platforms.
Ironically, the tightening of US sanctions and OpenAI's withdrawal from China will likely boost domestic AI development as Chinese firms profit from revenues that previously went to the US.
SenseTime's SenseNova and Alibaba's Tongyi Qianwen are seeing a rise in downloads and customer loyalty.
As Chinese developers make their multimodal features publicly available, one has to wonder how patient American users will proceed to be.
Will they wait for OpenAI and Google to maneuver from demo to product, or will we see American users adopt Chinese models?
