🚨 Info | OpenAI increases the speed of one of its models by up to 14x‼️‼️ OpenAI has announced the new Ultrafast mode 👀 With speeds reaching up to 750 tokens per second‼️ in collaboration with Cerebras ⚠️ The key point here isn’t just that responses will appear faster—AI agents performing dozens of sequential steps can complete long tasks much more quickly.

Zhipu releases GLM-5.3: Open-source models are now advancing from writing code to securing systems. Zhipu has launched GLM-5.3, whose focus extends beyond improved coding capabilities—it places cybersecurity at the core of the model’s mission. Built on a 743-billion-parameter foundation and further post-trained, GLM-5.3 specializes in intelligent programming and network defense. Compared to its predecessor, the company emphasizes achieving more complex tasks with fewer output tokens. This shift is significant: AI-powered programming is evolving from “helping you write a few lines of code” to becoming an autonomous agent capable of decomposing tasks, invoking tools, and executing processes continuously. Writing code faster is just the first step; the real differentiator lies in identifying vulnerabilities, managing permissions, and preventing minor errors from escalating into system-wide failures. GLM-5.3 is currently available through the GLM Coding Plan and ZCode, with API access and open weights to be rolled out in phases, alongside security evaluations. For open-source models, the next battleground is no longer just parameter scale—it’s who can simultaneously deliver capability, cost-efficiency, and security boundaries.

GPT-5.6 is 14x faster! 🚀 OpenAI has unveiled a new "Ultrafast" mode. The latest model, GPT-5.6 Sol, now operates 14 times faster, with a preview launched targeting enterprise users. The evolution of AI is accelerating. It’s exciting to see how it will be applied in business settings moving forward.

Holy sh*t. @Zai_org just introduced GLM 5.3. The cyber capabilities are insane, and scary. But this IS actually a massive opportunity for crypto protocols to audit their systems and protocols more efficiently and at a fraction of the cost. “Cyber Defense” indeed. Open source is the future.

Breaking: $CBRS powers OpenAI’s new GPT-5.6 Sol Ultrafast API tier at up to 750 output tokens per second, delivering up to 14× faster processing than Standard.

SN46 @instantsubnet might be one of the more interesting approaches to decentralized inference Instead of trying to solve the whole "can we trust the miner's hardware" problem with TEEs, they kinda flipped it around, miners provide the raw compute, while Instant runs its own code inside the miner setup through a virtual network runtime And the speed is kinda crazy too, 300 to 500 tokens/sec on open-weight models like GPT-OSS 20B. That's way faster than what you'd normally get from API providers like OpenRouter And apparently this is only the first step, the longer-term goal is getting into analog chip architectures where the model weights are basically built into the hardware If they can pull that off, decentralized inference on $TAO could get really damn fast

Cerebras just landed a major OpenAI win. $CBRS is powering the new Ultrafast tier for GPT-5.6 Sol, delivering up to 750 output tokens per second — as much as 14x faster than standard mode, according to the companies. The limited preview targets latency-sensitive work including financial research, incident response, security, voice and live experimentation. Early users include Jane Street, Basis and Rogo. The timing matters: Cerebras shares had fallen 13%+ after its Q2 results and guidance disappointed investors. Now the OpenAI partnership gives investors a fresh datapoint on demand for Cerebras’ inference infrastructure. Can AI inference speed become a meaningful competitive advantage for $CBRS?

What if AI latency stopped being the bottleneck? OpenAI is previewing Ultrafast mode for GPT-5.6 Sol, reaching up to 14× the speed of Standard processing. It’s launching in the API for select customers first, with wider access planned as capacity grows.