#codex

Microsoft Azure announced that Microsoft Foundry now offers GPT-Realtime-1.5, GPT-Audio-1.5, and GPT-5.3-Codex. The stated focus is low-latency voice interactions and long-running engineering workflows.

OpenAI and Figma announced a partnership that links Figma Make with OpenAI Codex workflows. The companies position the integration as a faster path from prompt and prototype to production-ready software.

OpenAI and Paradigm launched EVMbench, a benchmark for AI agent performance on smart contract detection, patching, and exploitation tasks. OpenAI reports GPT-5.3-Codex scored 72.2% in exploit mode versus 31.9% for GPT-5.

OpenAI and Figma launched a new integration that links Codex directly with Figma through an MCP-based workflow. The goal is to reduce context loss between implementation and design by enabling continuous code-to-canvas roundtrips.

OpenAI and Pacific Northwest National Laboratory introduced DraftNEPABench to evaluate AI coding agents on NEPA drafting tasks. In expert assessments, agents showed potential to reduce drafting time by 1 to 5 hours per subsection, or up to roughly 15%.

OpenAI announced GPT-5.3 Codex Spark on February 12, 2026, positioning it as a coding-focused model optimized for practical throughput and cost efficiency. The company reports lower latency and token cost versus GPT-5.2 while maintaining strong benchmark results.

A high-signal Hacker News discussion on GPT-5.3-Codex-Spark points to a shift toward low-latency coding loops: 1000+ tokens/s claims, transport and kernel optimizations, and patch-first interaction design.

Omnara’s Launch HN thread (February 12, 2026) reached 143 points and 153 comments. The discussion focused on remote continuity for locally running Claude Code/Codex sessions, plus tradeoffs around pricing, security boundaries, and alternatives like self-hosted workflows.

OpenAI launches GPT-5.3-Codex, the first model to debug its own training and manage deployment. Released with tight security controls due to cybersecurity concerns.

OpenAI and Anthropic engaged in a dramatic race to release their latest AI coding models on February 5, with Anthropic moving its announcement 15 minutes earlier and OpenAI responding with GPT-5.3-Codex, the first model that helped build itself.