OpenAI boosts GPT-5.6 with low-latency Ultrafast option amid industry push for specialised AI infrastructure

OpenAI introduces GPT-5.6’s Ultrafast mode in collaboration with Cerebras, signalling a competitive race in AI speed, hardware customisation, and supply chain realignment across tech giants including Apple, Google, and Microsoft.

OpenAI has expanded GPT-5.6 with a new Ultrafast option for the Sol tier, developed with Cerebras, as the company pushes its latest model family into lower-latency use cases. According to OpenAI, GPT-5.6 is being rolled out across ChatGPT, Codex and its API, with the Sol, Terra and Luna tiers aimed at different performance and cost needs. The company also says its Fast mode for API customers can deliver up to 2.5 times faster output, underscoring how model speed is becoming a competitive feature in its own right.

The move fits a wider industry shift towards specialised AI infrastructure. Major technology groups are investing heavily in proprietary chips, memory and data-centre capacity as they try to cut dependence on external suppliers and reduce energy use during inference. Google is developing new chips for Gemini, while Microsoft is preparing the next generation of Maia processors, reflecting a broader race to control more of the hardware stack behind advanced AI services.

That race is also reshaping supply chains. Apple is said to be under pressure from Washington over possible sourcing from Chinese memory makers, even as it searches for alternatives amid a global DRAM shortage driven by AI data-centre demand. Reporting from MacRumors and Tom’s Hardware indicates that Apple has been testing memory from CXMT and lobbying the US government for approval to use it, while lawmakers are urging restrictions on Chinese memory imports over national-security concerns. The company’s challenge is sharpened by a packaging shift at TSMC that reportedly leaves large numbers of completed Apple chips waiting for DRAM before final assembly.

Beyond chips and memory, the latest developments show how far the AI build-out now reaches. Anthropic has begun embedding invisible digital marks into Claude-generated text to support European transparency rules, while Chinese firm Z.ai is promoting cybersecurity capabilities in its open-source GLM-5.3 model. In robotics, Unitree’s expected Shanghai listing and the rapid growth of China’s humanoid-robot sector point to another frontier where the competition is no longer just about software, but also about scale, manufacturing and real-world deployment.

Disclaimer: This content is intended for informational purposes only. Readers are advised to exercise their own judgement, conduct due diligence, or consult a qualified expert before acting on any information provided.