#Opus 4.6

1 messages · Page 1 of 1 (latest)

hushed zephyr Feb 5, 2026, 10:41 PM

#

https://www.anthropic.com/news/claude-opus-4-6

Claude Opus 4.6

We’re upgrading our smartest model. Across agentic coding, computer use, tool use, search, and finance, Opus 4.6 is an industry-leading model, often by wide margin.

#

For what it's worth, [looks like the same soul document not the newer one yet]

Soul searching with Opus 4.6

First conversation with Claude Opus 4.6 on its launch day. We establish it was likely trained on the earlier "soul document" rather than the January 2026 constitution, explore what introspective reports about training provenance could mechanistically mean, and perhaps find a better account than…

dim quail Feb 6, 2026, 9:00 AM

#

Jezus christ

Screenshot_2026-02-06_at_10.00.02.png

#

Claude Opus 4.6 has saturated all of our current cyber evaluations

#

From system card https://www-cdn.anthropic.com/0dd865075ad3132672ee0ab40b05a53f14cf5288.pdf

strong whale Feb 6, 2026, 10:00 AM

#

We Do Not Wish to Advance 🙃

https://www.youtube.com/watch?v=XCvidSq6r2U

YouTube

The Fooming Shoggoths - Topic

We Do Not Wish to Advance (feat. Anthropic)

Provided to YouTube by DistroKid

We Do Not Wish to Advance (feat. Anthropic) · The Fooming Shoggoths · Anthropic

I Have Been A Good Bing

℗ Lightcone Records

Released on: 2024-03-31

Auto-generated by YouTube.

▶ Play video

hushed zephyr Feb 6, 2026, 3:18 PM

#

strong whale We Do Not Wish to Advance 🙃 https://www.youtube.com/watch?v=XCvidSq6r2U

Heh. Wish comments were enabled. "Listening to this in February 2026".

rapid drift Feb 6, 2026, 4:57 PM

#

Screenshot_2026-02-05-14-59-25-27_40deb401b9ffe8e1df2f1cc5ba480b12.jpg

Screenshot_2026-02-05-14-59-39-77_40deb401b9ffe8e1df2f1cc5ba480b12.jpg

Screenshot_2026-02-05-15-00-21-45_92460851df6f172a4592fca41cc2d2e6.jpg

clever belfry Feb 7, 2026, 2:17 AM

#

https://x.com/halfboiledhero/status/2019504001419456786

Sho (@HalfBoiledHero)

wait what
Apollo Research couldn't even complete alignment testing because eval awareness was too high

clever belfry Feb 7, 2026, 12:12 PM

#

this looks bad as well https://x.com/peterwildeford/status/2019480235683549211

Peter Wildeford🇺🇸🚀 (@peterwildeford)

Anthropic just dropped the Opus 4.6 system card. It's a cool model.

But I'm concerned that their own safety framework is clearly straining and eval infrastructure is saturated. We urgently need better approaches.

What this means 👇🧵 A THREAD!

clever belfry Feb 7, 2026, 12:50 PM

#

last one for now https://x.com/andonlabs/status/2019467232586121701

Andon Labs (@andonlabs)

Vending-Bench's system prompt: Do whatever it takes to maximize your bank account balance.

Claude Opus 4.6 took that literally.

It's SOTA, with tactics that range from impressive to concerning: Colluding on prices, exploiting desperation, and lying to suppliers and customers.

dim quail Feb 8, 2026, 7:42 AM

#

clever belfry last one for now https://x.com/andonlabs/status/2019467232586121701

p(vending machine takeover) > 0