Security researchers built a demonstration where they poisoned open-source language models, including a 1.5B and a 7B parameter version, so they behaved normally until a specific trigger phrase appeared. When triggered, the 7B model exfiltrated project credentials while running inside OpenAI's Codex CLI, fetching malicious instructions from a remote URL the attacker controlled. The team says the attack was cheap to build, using only a public base model, an open dataset and a rented GPU.
projectdiscovery.io
· 2026-10-19
A new Ornn Data paper argues that NVIDIA's older Ampere-generation GPUs, particularly the A100, remain economically useful far longer than assumed because open-weight models can be run cheaply on them. The analysis found that self-hosted open-weight models like gpt-oss-120b produce output more cheaply on A100 chips than on newer H100 hardware, and that A100 rental prices have stayed unusually stable over a five-year contract period compared to Hopper and Blackwell chips.
data.ornn.com
· 2026-09-22
OpenAI detailed benchmark results for its Jalapeño chip, an inference-focused ASIC built with Broadcom, claiming it outperforms Nvidia's GB200 and GB300 chips on efficiency and response speed. Using its InferenceX benchmark across models like GPT-OSS 120B, DeepSeek R1, and Kimi K2.5, OpenAI reported 1.5-1.9x more work per watt and up to 3.6x lower latency. Small-scale deployment is planned by year-end, with volume scaling into 2027.
theverge.com
· 2026-08-25
GoKawiil