GPT-6 Sol leaks, screenshots and release rumors

Codex CLI 0.156.1 adds GPT-6 Sol and GPT-6 Luna to the model picker and changes the rate-limit switch prompt to recommend GPT-6 Luna.

A leak claims OpenAI's GPT-6 Sol and Luna are priced 50% below GPT-5.6 promo rates, with Luna at $0.10 per million input tokens and $0.50 output, undercutting DeepSeek.

After using Fable, Astra and GPT-6 Sol, the author calls Opus 5.5 the strongest model on high reasoning and suggests a multi-model workflow for hard problems.

Episode 473 of a Japanese AI news channel reports the GPT-6 Sol & Luna release, the arrival of Opus 5.5 and the Qwen4 launch, and notes that Codex reset tickets have arrived.

An interactive question asks which model you would choose, Opus 5.5 or GPT-6 Sol, if you could use only one to build your next startup.

The post says GPT-6 Astra, Sol and Luna are not very good (a deliberately mild phrasing), prompting the view that users should move entirely to Claude.

A user says GPT-6 Sol's tone feels cold and calls it as bad as the '5.2 Karen bot,' hoping warmth matters more for everyday users when it launches in chat.

A user says they were very disappointed in GPT-6 Sol and pleasantly surprised by GPT-6 Luna, while Opus 5.5's generous limits and feel won them back after cancelling Anthropic.

Higgsfield posted a video comparing Claude Opus 5.5 and GPT-6 Sol building a mosaic wake-up animation in JavaScript.

A short Chinese post asks whether everyone has gotten access to gpt-6-sol's high fast mode, suggesting the model is already available to some users.

A long Chinese post breaks down GPT-6 Sol ($2 input/$10 output per million tokens) and Luna pricing, AutomationBench figures, and how the price war affects model selection and wrapper startups.

A meme imagining OpenAI responding to Anthropic's Opus 5.5 launch by immediately releasing GPT-6 Sol and GPT-6 Luna.