GPT-6 Sol leaks, screenshots and release rumors

The VulcanBench author says Luna at low reasoning effort declares jobs done in about three minutes without solving them correctly, with Extra High and Max results still pending.

A weekly AI recap covers GPT-6 Sol and Luna launching the same day as Claude Opus 5.5, API pricing 50 percent below GPT-5.6 promo rates, plus OpenAI math claims, misalignment incidents and other news.

TokenGremlin argues Sol is essentially a heavily upgraded GPT-5.6 that picked up a few tricks from Astra, which explains why it loses to Opus 5.5 and Sonnet 5.5.

TokenGremlin says OpenAI keeps squeezing the old Doug pretraining base instead of building on the newer Bel, leaving Sol worse than GPT-5.6 Sol on some benchmarks with price as its main advantage.

Anthropic said at the Opus 5.5 launch that Sonnet 5.5 and Haiku 5.5 would follow within weeks, leakers point to Sept 30 or Oct 1, and the author says the circulating "Sonnet 5.5 beats GPT-6 Sol" screenshot is faked from old Sonnet 5 numbers.

AdamHoltererer lists DevDay predictions: GPT Live 2, GPT-Image 3, Astra 6.1, a Bel preview, and a 700 tokens-per-second ultrafast mode for GPT-6 Sol.

xAI cut prices first, then Anthropic and OpenAI followed, with Sol at $2 in and $10 out and Luna as low as 10 cents per million tokens, which the author reads as the intelligence race stalling into a race to the cheapest.

MiniMax's new model appeared quietly in MiniMax Code, and the author says it is very fast and posts a speed comparison against GPT-6 Sol at max reasoning.

The author asked Claude Opus 5.5 and GPT-6 Sol to create the same glitter sticker effect and concludes Opus 5.5 is clearly stronger on this task, with a video comparison.

The user complains that GPT-6 Sol and the smaller models of the same generation are extremely dumb, saying computer use is good but reasoning is poor, and that Opus 5.5 wins.

After testing, the user says GPT-6 Sol high does not beat the previous generation's GPT-5.6 Sol xhigh, calling it the first time a big version bump brought barely any improvement.

Comparing paper writing, review, search, readability and slide design, the author finds Opus text easier to read and GPT more rigorous, and decides to keep Pro Lite.