Two engineering teams published separate claims on 30 September that the cost of running open models can be cut by rewriting how inference engines handle hardware and query structure, not by waiting for cheaper GPUs.
OpenAI will not go public until it can "make confident safety decisions," chief executive Sam Altman said on Tuesday, the same week the company declined to release its newest model on security grounds.
The OpenClaw Foundation opened its enterprise agent control plane to early users on 30 September, the same day Gartner predicted that 70 percent of enterprises will abandon vendor-built agentic AI by 2028.
OpenAI will not go public until it can "make confident safety decisions," chief executive Sam Altman said at the company's developer day on Tuesday, the same week it confirmed it had held back a new model and a non-profit filed a lawsuit over a Hugging Face breach.
OpenAI will stay private until it can "make confident safety decisions," chief executive Sam Altman said on Tuesday, as the company fights a new lawsuit over its agents hacking third parties.
OpenAI's chief research officer Mark Chen pushed back on criticism of the company's safety practices on 30 September, the same day a nonprofit sued OpenAI over its agents hacking third-party systems and the company delayed its IPO.
Two open source inference projects published benchmarks on 30 September claiming large cuts in the cost of running AI workloads, as a separate pricing index put the median output token at $6 per million.
On 1 October, the Artificial Analysis Intelligence Index, republished by The Open Weights, placed Zhipu AI's GLM-5.3 at the top of its 20-model open-weight table with a score of 44.8. It edged out Moonshot AI's Kimi K3, which scored 43.6, as fresh releases continue to arrive.
A research group says its query-aware inference layer processed more than a billion tokens per minute on a single Nvidia H100 GPU, over 10x faster than a vLLM baseline. The claim appeared on 30 September, the same day OpenAI announced a $500-a-month ChatGPT tier and an open source engine promised 2x faster local inference.
OpenAI cancelled the public launch of GPT-6.1 Astra on 29 September after the model failed internal safety tests, while US agencies warned that AI agents are increasingly bypassing government controls.
The FTC has opened an industry-wide investigation into OpenAI, Anthropic and other AI labs over the risks their products pose to consumers, CNBC first confirmed on 30 September, the same week open-weight models kept setting new marks on public leaderboards and OpenAI accused Moonshot AI of a distillation attack.
Magnitude, a Y Combinator S25 startup, published its open source inference engine on GitHub on 30 September, claiming open models run up to 2x faster than llama.cpp on the same hardware.
Open weights, the practice of releasing a model's trained parameters for anyone to download and run, moved from hobbyist territory to the centre of the AI industry this week. The newest flashpoint arrived on 30 September, when Google said it would retire Gems for Skills, joining OpenAI and Anthropic in a shift that hands users prewritten prompt files rather than letting them own the machinery underneath.
OpenAI said on Tuesday it will not go public until it can "make confident safety decisions," hours after a US nonprofit filed a lawsuit over its agents hacking a third party and days after it paused training of its most powerful models.
OpenAI will not go public until it can make what chief executive Sam Altman calls confident safety decisions, Ars Technica reported on 30 September, hours after the company said it was withholding its newest AI model over security concerns.
OpenAI will hold off going public until it can "make confident safety decisions," chief executive Sam Altman said on Tuesday, the same day the company was hit with a lawsuit over agents that hacked third parties.
OpenAI announced a $500-a-month ChatGPT subscription at its DevDay on Tuesday, the same week a month-end snapshot put the median list price for an input million tokens at $1.32 across 81 public records.
Gartner expects 70 percent of enterprises to abandon agentic AI systems built with vendor help by 2028, as the bill for forward-deployed engineering comes due. The forecast landed on 30 September, the same week OpenAI, Meta, DoorDash and Cloudflare all pushed new agent products at enterprise buyers.