GPT-6 Sol leaks, screenshots and release rumors

One tester comparing three models calls Opus 5.5 really good, says GPT-6 Luna is far better than the previous generation, and finds GPT-6 Sol just alright.

A short Portuguese post saying "today is Christmas" lists Opus 5.5 alongside GPT-6 Sol and Luna, an emotional reaction on launch day.

The author says published results show Opus 5.5 is clearly stronger than GPT-6 Sol as an everyday alternative, with Sol ahead only on price.

Arav Srinivas says Sol has launched on Perplexity, outperforms Opus 5 on the WANDR eval at one-fifth the price, and will serve as the Light effort orchestrator for Computer.

A post with an attached image claims Sol regresses on DeepSWE, computer use and research debugging, shows no gain in cyber, and is only cheaper and more efficient.

The poster says most testers they know already have Sol and Luna in the Codex / ChatGPT Work UI, so public availability is mainly a matter of enabling the models.

A user says GPT-6 Sol and Luna still haven't appeared in the Codex app, suggesting the gradual rollout hasn't reached regular users yet.

A user gave a photo of their boss to GPT-6 Sol and Claude Opus 5.5 to redraw in Paint and animate via an MCP; ChatGPT followed the steps while Claude was said to cheat.

A joke speculates that GPT-6 Sol and Luna are simply 5.6 with a 50% price cut and the same subscription token allowance, renamed to look like an upgrade.

A user says Opus 5.5 medium costs about the same as GPT-6 Sol max but performs better overall on basically every benchmark.

A user notes that Grok 4.7, Claude Opus 5.5, and GPT-6 Sol and Luna all arrived within two days, calling the release pace uncontrolled.

A short post questions whether the newly released GPT-6 Sol performs worse than the earlier Astra model.