GPT-6 Sol leaks, screenshots and release rumors

Four models each generated three.js Star Wars scenes from one prompt, judged by automated headless Chrome screenshots; Sol was fastest on all three tasks at a total cost of $6.99.

A Japanese post collects an overseas user's method for dividing work among the three models: Astra for multi-step jobs, Sol for complex reasoning and agent work, Luna for repetitive extraction, plus token-saving trade-offs.

MacRumors says OpenAI's new GPT-6 Sol and Luna models bring Astra's improvements to lower-priced subscription tiers.

A chart compares Anthropic's Opus 5.5 with GPT-6 Sol and Luna on price, context, AA index and cost per task.

The user argues Opus 5.5 already beats Sol, that Sol and Luna only cut costs without quality gains, and that OpenAI needs Astra 6.1 to catch up.

A comparison video asks viewers to guess whether the edit was made by Claude Opus 5.5 or GPT-6 Sol.

A user reports that switching from Astra xhigh to Sol xhigh cut weekly quota use from 5-7% to 1% for the same workload.

A roundup post says that after last week's talk of an AI slowdown, OpenAI released GPT-6 Sol and Luna, alongside Opus 5.5, Grok 4.7 and Meta Muse.

Raycast's maker says both models are live, with Sol offering twice the accuracy at half the cost and Luna positioned as an efficient model for everyday tasks.

The Synara changelog says GPT-6 Sol and Luna are now in Codex, GPT-6 Luna takes over commit messages and PR text, and Astra remains the default for chat.

A video title says OpenAI released ChatGPT-6 Sol and GPT-6 Luna as a response to Anthropic's launch.

A recap of Anthropic and OpenAI launching on the same day: Opus 5.5 beats Fable 5.1 on every benchmark and costs about 40% less, while GPT-6 Sol nears Astra reliability and Luna is very cheap.