Advancing the price-performance frontier with GPT-5.6
OpenAI has reduced pricing for GPT-5.6, lowering the cost barrier for running enterprise-scale agentic workflows.
135 candidate events; 40 shortlisted stories; 12 selected. Counts describe different stages; clustering, validation and refills can change the shortlist.
Active filters: none. Sort: Ranked.
All 12 stories, in ranked orderOpenAI has reduced pricing for GPT-5.6, lowering the cost barrier for running enterprise-scale agentic workflows.
Gemini Robotics 2 expands from upper-body manipulation to full whole-body control for humanoid robots.
Enabling reasoning retention and compaction API settings tripled GPT-5.6 scores on the difficult ARC-AGI-3 benchmark.
NVIDIA details how identical H100 and GB200 clusters yield different training throughputs due to infrastructure configuration.
A new open-source engine runs the 26B Gemma 4 model locally on M-series Macs using only 2GB RAM.
Updates to the Model Context Protocol (MCP) establish stability policies to prevent sudden feature deprecation in enterprise setups.
Tencent's open-source AngelSpec framework enables training of speculative-decoding draft models across six different architectures.
New research presented at ICML mathematically demonstrates that LLMs cannot be fully secured against adversarial prompt injection.
Tokenless automates model switching dynamically to optimize API costs without manual routing logic.
Researchers demonstrated self-propagating document-borne AI worms that exploit Copilot for Word to infect other files.
The GCC steering committee has officially established its policy on integrating AI-generated code into the compiler.
Okta's $200M acquisition of Permiso targets the growing need to secure non-human identities and autonomous AI agents.
No stories match. Active filters.