Social Media Briefing — June 27, 2026

Following yesterday's News, OpenAI outlines GA plans, OpenAI announces it will make GPT-5.6 Sol, Terra, and Luna generally available in coming weeks but begins with a limited preview among trusted partners in Codex and the API at US government request.

Following yesterday's News, OpenAI details the lineup, OpenAI positions Sol as a step-function flagship, Terra as 5.5-level at half cost, and Luna as most cost-efficient, giving users choice across intelligence, speed, and cost.

Natolambert critiques sloppy thinking on banning open models, arguing bans won't stop global open-model progress or bad actors, questioning what is gained by banning models including Chinese ones.

Chollet argues that benchmarks based on static datasets measure memorization and retrieval, not intelligence, and should not be conflated with it.

Following yesterday's News, OpenAI touts Sol's security capabilities, OpenAI says GPT-5.6 Sol is its most capable cybersecurity model, advancing long-horizon security tasks including vulnerability research and exploitation.

OpenAI claims GPT-5.6 Sol sets a new state of the art on Terminal-Bench 2.1 for complex command-line workflows.

OpenAI states GPT-5.6 Sol launches with its most robust safety stack, citing strengthened real-time protections against cyber misuse and over 700,000 A100-equivalent GPU hours of automated red teaming.

vLLM announces support for serving NVIDIA's official NVFP4 checkpoint of GLM-5.2 on Blackwell, claiming reduced memory footprint vs FP8 while matching accuracy on reasoning, coding, and long-context benchmarks.

Argues that a Claude enterprise tagging product creates deep vendor lock-in by absorbing company data, urging open source to preserve flexibility to swap models and harnesses.