GPT-6 Sol leaks, screenshots and release rumors

An Italian-language post says Opus 5.5 is a tier ahead of GPT-6 Sol and the two models cannot currently be compared.

A short review says Claude Opus 5.5 performed impressively while GPT-6 Sol had little to offer beyond a low price.

A user tested GPT-6 Sol at MAX effort on the $20 Codex plan: 15 million tokens and 7% of the weekly limit in 35 minutes, but the run was very slow and did not complete the task.

A Sept 23 brief says GPT-6 Sol and Luna rolled out in ChatGPT Work and Codex, with API prices 50% below the GPT-5.6 promo, alongside news from Anthropic, SpaceXAI, Cursor, Perplexity and DigitalOcean.

Community test results put the API sweet spot between GPT-6 Luna high and GPT-6 Sol max, with Sol xhigh offering the lowest marginal cost per point of intelligence.

A poster claims GPT-6 Sol will be added to Chat mode with unlimited use and says OpenAI staff confirmed it, sharing a screenshot.

A video shows that after signing in at Experiential Labs, users can call gpt-6-sol, gpt-6-luna and Claude Opus 5.5 for free from any OpenAI-compatible client.

flowith posts a video comparing the outputs of Claude Opus 5.5 and GPT-6 Sol given the same draft request.

Taskade ran Claude Opus 5.5, GPT-6 Sol and GPT-6 Luna on the same real tasks in its TSK-1 benchmark, with early scores of 8/10, 7/10 and 5/10.

A user shares an AA intelligence index screenshot and says Muse Spark 1.3 still beats GPT-6 Sol, Grok and every Chinese AI lab, predicting Meta will overtake OpenAI within months.

A Spanish-speaking user says over an hour of writing with GPT-6 Sol used none of the weekly quota, calling it efficient and as good as 5.6 for their tasks.

A comparison question post asks how users would choose between GPT-6 Sol and Astra, with replies likely reflecting how the two are received.