GPT-6 Sol leaks, screenshots and release rumors

The author tests four models with one news-writing prompt, lists costs and results, and calls Opus 5.5 the best result and GPT-6 Sol the cheapest.

The poster compares Opus 5.5 with GPT-6 Sol and Luna, noting Sol only shipped in Work and Codex and appears to be an optimized version of 5.6.

A Spanish-language post says OpenAI launched GPT-6 Sol and Luna amid calls to pause AI, with frontier model API prices down as much as 50%.

The author, finding benchmark scores and token prices unintuitive, gathers overseas users' real-world outputs from running identical prompts on Opus 5.5 and GPT-6 Sol.

A screenshot post claims Meta's Muse Spark 1.3 Max is better, cheaper and faster than GPT-6 Sol, as a third-party model comparison.

A weekly roundup says GPT-6 Sol and Terra are roughly 50% cheaper and faster, Opus 5.5 is about 40% cheaper and noticeably better at writing, and also mentions JEV, Muse and Grok 4.7.

pi editor v0.87.1 adds Opus 5.5 and GPT-6 Sol/Luna models, noting both have a 200k context so users wanting 1M must edit the config file manually.

A tester says Sol does worse than 5.6 on some coding tasks, with the halved price as the main change, while Astra burns through limits in minutes.

An investment note says OpenAI released the budget GPT-6 Sol and Luna with permanent price cuts and Perplexity set them as default, while Anthropic shipped Opus 5.5 the same day, arguing value is shifting to compute infrastructure.

The user says Opus 5.5 is strong at Japanese writing, reasoning and creative work at reasonable cost, while GPT-6 Sol brought no noticeable change.

An updated price-performance curve chart calls GPT-6 Luna the best at the low end and Opus 5.5 the best value with frontier performance, while noting Fable 5.1 is still preferred internally.

A repost says GPT-6 Sol has launched and describes building a 3D game in Codex with Astra planning and multiple Sol subagents, noting character modeling and large scene assets remain weak.