GPT-5.6 about to DROP

Video opens with highlight reel preview — Just Vibes (50/100)

At 0:00

Bold preview — let's see if the video delivers on these promises.

Why this score: Rapid teaser montage listing multiple unproven future events. Typical hype framing without supporting details yet.

Original quote: “There's a few big stories in AI. Let's get started. So, Anthropic is moving closer to an IPO. We might have for the first time in history $3 trillion IPOs all in the same year. That's SpaceX, Anthropic, and OpenAI. We're going to find out if this is an AI bubble or not one way or another very, very…”

Anthropic filed confidentially for IPO at nearly $8 trillion valuation — Sketchy (20/100)

At 0:30

Re-scored to 20/100 after evidence check.

Why this score: Sources contradict this claim. [San Francisco AI giant Anthropic files for IPO after $965 billion valuation](https://www.sfchronicle.com/tech/article/anthropic-confidential-ipo-filing-22286123.php), [Anthropic files to go public in a potentially trillion-dollar debut | CNN Business](https://www.cnn.com/2026/06/01/tech/anthropic-ipo-filing).

Original quote: “multiple outlets are reporting that Anthropic filed confidentially for a US IPO... Their valuation is just under $8 trillion. The latest one was $965 billion.”

Sources: San Francisco AI giant Anthropic files for IPO after $965 billion valuation, Anthropic files to go public in a potentially trillion-dollar debut | CNN Business, The AI IPO race is on: Anthropic confidentially files after surpassing OpenAI in valuation - San Francisco Business Times

Opus 4.8 built entire simulation in ultra mode — Unverifiable (50/100)

At 4:30

No public evidence Opus 4.8 exists — naming feels off.

Why this score: Unverifiable model name. - No records of Opus 4.8 or "ultra code" mode in current releases - Speaker provides no screenshots or reproducible steps - Web results mention GPT-5.5/5.6 speculation only

Original quote: “This was entirely built by Opus 4.8 in their ultra mode.”

Claude Opus 4.8 loses to GPT-5.5 on software engineering benchmark — Unverifiable (50/100)

At 6:32

No benchmark name, numbers, or source given — pure assertion

Why this score: No source provided. - Speaker references a chart but names no benchmark (SWE-bench, AgentBench, etc.) - No methodology or sample size mentioned - Web context shows zero independent verification of these specific scores

Original quote: “We're seeing that cloud opus 4.8 is not as good as GPT 5.5. Neither the extra high model nor the max model outperform GPT 5.5.”

Benchmark skipped Claude's 'ultra code' mode entirely — Opinion (50/100)

At 6:38

Fair critique of missing test condition — still speculation

Why this score: Legitimate point on methodology. - Speaker correctly notes the benchmark omitted the highest effort setting - Without that data, direct comparison remains incomplete - No evidence either way on whether it would close the gap

Original quote: “Both high and extra high are above clot opus 4.8. But here's the thing. They measured the extra high. They measured the max. So here's the extra high. Here's the max. Here's the problem. For whatever reason, they did not include ultra code.”

If Claude ultra code loses, Anthropic is falling behind OpenAI — Opinion (50/100)

At 6:48

One benchmark ≠ company trajectory — massive leap

Why this score: Over-extrapolation. - A single coding benchmark does not define overall model leadership - Web context shows ongoing speculation about GPT-5.6 release, not confirmed dominance - Real competition involves multiple domains, not just software engineering scores

Original quote: “I would have loved to see how well that ultra code performs here because if it falls flat, if GPT 5.5 still is the ruling coding model, well, that would definitely be pointing to the fact that maybe Anthropic is falling behind OpenAI.”

RKGI benchmark tests fluid intelligence in AI models — OK (55/100)

At 8:30

Sounds plausible but no public info on what RKGI actually is.

Why this score: No verifiable source for an RKGI benchmark. Key issues: - No academic papers or public leaderboards reference it - Speaker doesn't cite any external validation - Could be an internal or renamed test

Original quote: “the RKGI, it fundamentally tries to test something different. It's trying to test fluid intelligence.”

Opus 4.8 set SOTA on ARC-AGI 3 at 1.5% — Dubious (40/100)

At 8:39

ARC-AGI 3 leaderboard doesn't list any Opus 4.8 model at 1.5%.

Why this score: No matching result on official ARC-AGI sources. Key issues: - Current public leaderboards show top scores still below 1% - No announcement from ARC Prize team about Opus 4.8 - Model naming (Opus 4.8) doesn't match known Claude releases

Original quote: “Opus 4.8 8 has been a massive jump on the ARC AGI 3. So the ARC prize, the people behind ARGI, they're saying that Opus 4.8 reached the state-of-the-art performance on ARGI 3.”

Opus 4.8 scored 1.5% on some benchmark while others sit at 0.5% or below — Unverifiable (50/100)

At 10:30

No benchmark named — can't check the actual numbers

Why this score: No source provided for the scores. - Speaker references an unnamed benchmark with "clot opus 4.8" at 1.5% - Other models described as 0.5% or below - No link, paper, or leaderboard cited to verify these exact figures

Original quote: “most of them are at 0.5% or below. And of course, clot opus 4.8 sits at 1.5%.”

GPT-5.6 and GPT-5.6 Pro spotted in OpenAI Codex backlogs, possible release tomorrow or Tuesday — Dubious (45/100)

At 12:30

Rumors are real but evidence is thin — no official confirmation anywhere.

Why this score: Rumors exist but lack concrete proof. - Multiple Reddit threads show people speculating about Codex sightings - No OpenAI announcement or benchmark release as of June 2, 2026 - Speaker correctly notes "nothing official" yet still amplifies the noise

Original quote: “there's some rumors swirling around about the potential release of GPT 5.6. So again, when we're talking about rumors, you know, take it with a grain of salt. Nothing is confirmed yet, but a lot of people are saying that they're beginning to see references to GPT 5.6 in the OpenAI codeex backlogs…”

OpenAI released GPT-5.5 in April, so GPT-5.6 coming in June means frequent rolling updates — Sketchy (25/100)

At 14:30

GPT-5.5 in April? No evidence of that release either — timeline is invented 🗓️💀

Why this score: Timeline doesn't match reality. Key issues: - No records of a GPT-5.5 April release exist in current web data - Reddit discussions treat GPT-5.6 as hypothetical, not following a confirmed 5.5 - Assumes a release cadence that hasn't been demonstrated yet

Original quote: “GPT 5.5 is released by OpenAI in April. So if they're releasing a GPT 5.6 six in let's say June that really suggest these Frontier model launches. They're not going to be these once a year type of events”