GPT-6 Sol leaks, screenshots and release rumors

Higgsfield compares GPT-6 Sol and Claude Opus 5.5 on building a 3D strategy game prototype at battlefield scale.

The author says Opus 5.5 cut costs to two-fifths and ran twice as fast with no quality loss, while GPT-6 Sol is faster and cheaper but less accurate than GPT-5.5.

A demo shows GPT-6 Sol or Opus 5.5 in Codex or Claude Code installing Xcode 27.1 beta, setting up the iPhone Duo simulator, and building and launching an app.

Higgsfield has both models build chainmail node networks in Houdini and test collisions, comparing GPT-6 Sol against Opus 5.5.

With the same prompt and scene, Sol cost about $34.55 and ran faster but had rigging and geometry issues, while Opus 5.5 cost $48.11, ran slower, and produced a more polished result.

One user says the model naming is confusing and suggests renaming GPT-6 Sol to GPT-6 Terra, with Astra Minor taking the Sol name.

A user notes that GPT-6 Sol costs less per token than Opus 5.5, yet on $20 plans it seems to use up Codex limits faster than Opus 5.5 uses Claude Code limits, and asks why.

A screenshot says ChatGPT Voice can now use email, calendar and Slack plugins, run on GPT-6 Astra, Sol and Luna, and create docs, spreadsheets and websites by voice.

OpenAI's Voice Mode update adds plugins for email, calendar and Slack, is powered by GPT-6 Astra, Sol and Luna, and works inside ChatGPT Work for larger tasks.

A post says GPT-6 Sol is very fast and uses fewer tokens, letting a half-formed game concept become a playable hovercraft racer now available on Combos Fun.

A user asks why GPT-6 Sol is so lackluster if OpenAI's models are as miraculous as claimed, saying something doesn't add up.

A poster says Claude Opus 5.5 and GPT-6 Sol have both shown up on the VoxelBench leaderboard, with a screenshot attached.