By the TensorMax editorial team
· Drawing from sources across the AI industry
Today's top story
model release
$40
Jev was announced by Diogo Almeida, a former OpenAI employee and co-inventor of reinforcement learning from human feedback, on September 15th.
Why it matters. The release of Jev, a probabilistic classifier optimized for fast, structured decision-making, has significant implications for the AI industry. With the ability to make decisions in 70-500 ms at a cost of $0.042 per million input tokens, Jev has the potential to disrupt the market by providing a cheaper and more efficient alternative to large language models. TypeSafe AI, the company behind Jev, has raised $40 million in seed funding and is positioning Jev as a key component in the development of agentic AI, which could lead to significant advancements in areas such as software automation and machine-to-machine processing.
Jev was announced by Diogo Almeida, a former OpenAI employee and co-inventor of reinforcement learning from human feedback, on September 15th. The model is designed to provide probabilistic calibrated decisions, rather than autoregressively predicting the next token, and has been described as a System-1 model. TypeSafe AI claims that Jev is up to 100 times faster and 100 times cheaper than conventional large language models for certain tasks. The model has already gained significant interest from developers, with popular model gateway platforms such as Vercel, Cloudflare, LangChain, and Langfuse adding Jev to their stacks. Jev is anticipated to be useful in agentic AI tasks like model routing and has the potential to unlock new possibilities for AI development. With its release, Jev has sparked debate about the future of AI and the potential for specialized models to complement or replace large language models. As the AI industry continues to evolve, Jev is likely to play a significant role in shaping the direction of AI research and development. The model's efficiency and cost-effectiveness make it an attractive option for developers looking to incorporate intelligence into their code, and its potential applications are vast, ranging from software automation to design and research paper classification. As of September 22nd, TypeSafe AI's Jev has been overwhelmed with signups, with the company pausing signups due to high demand.
More from today
product launch
Why it matters. Alibaba's debut of China's most powerful AI chip marks a significant strategic stake in the country's AI development, with the company aiming to train an AI model with up to 10 trillion parameters. This move underscores Alibaba's commitment to building end-to-end AI capabilities, with a focus on creating value through technological breakthroughs. Notably, the company plans to scale its Alibaba Cloud global data-centre capacity to over 20 gigawatts by 2032, highlighting the substantial investment and resources being dedicated to this endeavor.
model release
Why it matters. Alibaba's plans to train an AI model with up to 10 trillion parameters and debut China's most powerful AI chip underscore the company's strategic stake in the AI market, with a focus on building end-to-end capabilities to drive value creation. With a goal to scale its Alibaba Cloud global data-centre capacity to over 20 gigawatts by 2032, Alibaba is poised to significantly enhance its computing power and support the development of increasingly complex AI models, potentially giving it a competitive edge in the market.
model release
Why it matters. The release of Meta's new AI agent, Muse, has significant implications for the AI market, with over 902,000 downloads in just six days, surpassing Meta AI's 773,000 downloads in the same period. This rapid adoption poses a threat to competitors, particularly Amazon, which has blocked Muse from accessing its platform, citing security risks and terms of service violations. As Muse continues to gain traction, its impact on the market will be substantial, with potential consequences for other AI companies and the future of consumer AI assistants.
model release
Why it matters. The solving of 100 open problems in math by OpenAI marks a significant milestone in the development of artificial intelligence, demonstrating the capability of AI systems to tackle complex mathematical challenges. This achievement has the potential to impact various fields, including science, engineering, and finance, where mathematical modeling is crucial. With 100 problems solved, OpenAI's AI system has shown its ability to process and analyze vast amounts of mathematical data, making it a powerful tool for researchers and scientists. This breakthrough could accelerate advancements in multiple disciplines, as AI-assisted mathematical problem-solving becomes more prevalent.
model release
Why it matters. Xiaomi's release of the open-source MiMo-V2.6 Pro and Flash models marks a significant milestone in the AI industry, with the Pro model allegedly performing on par with Opus 5 and GPT-5.6 Sol across most agent benchmarks. This development has the potential to disrupt the market, as the MiMo-V2.6 Pro model scores 46 on the Artificial Analysis Intelligence Index, making it the top-scoring open-weight model, and is available at a significantly lower cost, with a price of $0.13 per Intelligence Index task.
benchmark result
Why it matters. The recent surge in downloads of Meta's Muse AI agent, with over 902,000 downloads in six days, surpassing ChatGPT's launch period, has significant implications for the AI market. This milestone indicates a substantial interest in personal AI agents, and Meta's success in this area could potentially disrupt the market, particularly with its partnership with Shopify to enable agentic checkout. The fact that Amazon has blocked Muse from accessing its platform adds to the complexity of the situation, highlighting the ongoing battle for dominance in the AI-powered shopping landscape.
Catch up quick
-
OpenAI's 10,000-agent swarm achieves a significant breakthrough in solving the Navier-Stokes problem, outperforming previous models
-
Cisco Talos researchers develop CAIRN, an open-source framework to track and analyze AI-integrated malware, and discover a fully autonomous hacking tool called CLOSEDQUORUM
-
Grok triples its score on Terminal-Bench 4.0, overtaking GPT-5.6 Sol with a score of 38.0%
-
Meta invests $14.3 billion in data-labeling company Scale AI
-
Shopify partners with Meta to allow Muse AI agent to complete purchases on behalf of users
-
Meta's stock price has increased by 10% following the release of its new AI agent, Muse.
-
Meta's stock price jumps 12% after Muse AI assistant downloads surpass 902,000 in six days
-
Alibaba plans to train a 5T- to 10T-parameter AI model as part of a sweeping push across AI models, chips, and data centers
-
Microsoft takes down EvilTokens AI-chatbot for cybercriminals in collaboration with law enforcement
-
British Columbia sues OpenAI over its role in a mass shooting
-
The state of California sues Tesla for allegedly fostering racial discrimination at its Fremont factory, seeking financial damages and injunctive relief.
-
Ukrainian naval drone sinks Russian kamikaze drone boat in the Black Sea
-
Serbian military debuts robot dogs, potentially built with Chinese technology, in large-scale tactical exercise
-
Chinese military doctors develop treatments for drone-inflicted head injuries, citing lessons from Ukraine war
-
China breaks ground on the world's first supercritical CO2 and molten salt energy plant in Shandong province
Also on the desk
- Meta's Muse AI agent has a zero-day vulnerability that can be exploited by attackers
- Researchers discover Confused Deputy Attacks vulnerability in AI Accelerators from Google, NVIDIA, Hailo, Texas Instruments, NXP, AWS, and Rockchip, impacting over 128 System-on-Chips and 100 million devices
- Alibaba plans to train a 5T- to 10T-parameter AI model as part of a sweeping push across AI models, chips, and data centers
- Larry Ellison's Oracle and the Ellison family gain control of CNN, CBS, HBO, Paramount, Warner Brothers, and other properties through a $110 billion merger
- Over $7 trillion to be spent on scaling AI over the next five years, according to Goldman Sachs
Cutting-room floor.Hit reply with one thing you'd delete from today's brief. Filler, repeated beats, a story you skimmed — that's the feedback we act on.
|
|