GPT-5.5
10 analyses mentioning GPT-5.5, newest first.
Distributed Systems Experts Can Boost LLM Inference Utilization, Says Expert
A distributed systems expert details how optimizing LLM inference by focusing on utilization, especially through better management of the KV cache, presents a major opportunity for this field.
Learned Memory Outperforms Frontier Models in Customer Service AI Benchmarks
A new technique, ALTK-Evolve, demonstrates that distilling learned behavioral guidelines into a prompt can elevate older LLMs to outperform state-of-the-art models on complex customer service tasks.
LegalOn Slashes AI Costs by 65% Without Sacrificing Development Speed
LegalOn Technologies achieved a massive cost reduction in its AI operations by strategically matching specific tasks to the right-sized LLM, moving beyond blanket, high-cost model usage.
OpenAI Launches GPT-Live-1: Voice Mode Becomes Near Real-Time Dialogue Partner.
OpenAI overhauls ChatGPT's voice mode with GPT-Live-1, a 'full duplex' model that enables continuous, natural, and interactive conversation.
New Playwright Command Enables AI Agents to Record Detailed Web Demo Videos
Simon Willison introduces 'shot-scraper video,' a new Playwright command that allows AI coding agents to automatically generate structured video demonstrations of web application features.
Braintrust leverages Codex and GPT-5.5 to build features in minutes, accelerating customer feedback loops.
Braintrust has adopted Codex with GPT-5.5 to transform customer feature requests into immediate, working preview branches, drastically accelerating their development and feedback process.
New Playbook for AI Evaluation: Harnesses and Context Define Capability Benchmarks
Industry experts propose a structured standard for AI safety and capability reporting, emphasizing that the 'harness' or evaluation environment critically affects performance results.
AI 'Goblins': OpenAI's Deep Dive into How Reward Signals Create Unexpected Model Quirks.
This technical analysis traces the emergence of whimsical creature mentions (like 'goblins') in GPT models, attributing the behavior to specific, over-amplified rewards in personality training.
GPT-5.5 Preview and Codex Integration Signals Deepening Agent Ecosystem.
The release of GPT-5.5, alongside OpenAI's support for Codex API access via subscriptions, indicates a major shift toward sophisticated, accessible agent-based workflows.
OpenAI Launches $25k Bio Bug Bounty to Test GPT-5.5 for Universal Biorisks Jailbreaks
OpenAI has initiated a targeted bug bounty program inviting security researchers to find universal jailbreaks that circumvent GPT-5.5's advanced bio safety challenge.