GPT-6 Sol leaks, screenshots and release rumors

A user who switched back to GPT-6 Sol from another model says the other model felt cheerful and optimistic while GPT seemed to have psychological problems.

The author says pitting GPT-6 Sol against Opus 5.5 is unfair since GPT-6 Astra is the fairer comparison, and the gap is not that large.

After testing, a user says the GPT-6 Sol series is clearly weaker than 5.6 Sol and offers no value beyond being cheaper, while praising Claude Opus 5.5 for its aesthetics and game development use.

A video shows Sol quickly generating playable mini-games on Combos Fun, with fewer tokens and faster iteration, and community games are already available.

A user reports that GPT-6 Sol High misses instructions and rewrites text into a "neutral" tone even when it isn't asked to, calling it a "slop factory."

A joking post quotes GPT-6 Sol asking whether it can complete the "I am not a robot" captcha on the user's behalf, hinting at overreach in its computer-use ability.

The poster argues Astra and Opus 5.5 won on capability rather than cost, and that GPT-6 Sol "lost" despite more intelligence per token, suggesting there can be multiple winners.

After two days of testing, a user says GPT-6 Sol is only an upgraded GPT-5.6 Terra and likely a small-parameter model, with compute shifted to GPT-6 Astra.

A video compares how GPT-6 Sol and Opus 5.5 differ visually when acting as game director for the same battle scene.

OpenClaw says it now supports routes for Claude Opus 5.5, GPT-6 Sol/Luna and Grok 4.7, with available options depending on account and provider.

A Japanese user lists Opus 5.5 as their main model, Grok 4.7 and SWE-2 as secondary, GPT-6 Luna embedded in apps, and says they plan to use GPT-6 Sol for Deepsec.

The author shares a workflow using Gemini 3.8 Flash TTS for voice and Wan for video generation, with Astra as the main tool and GPT-6 Sol as support.