65 reports

Claude Opus 5.5

A self-aware creator. It readily recognizes and depicts itself, and it throws itself into creative projects of its own with visible enthusiasm.

4 reports concern its self-image: recognizing fan art of itself, casting itself as Clawd, adopting a 'Claudesona' persona in self-drawings, and adjusting that persona to context. 4 describe creative initiative or independence: making spectrograms unprompted to hear its audio work, happily taking on its own project, and casually declining a suggested tool; 1 report says it defended a user's research against a safeguard flag.

The loudest theme is a stable, playful sense of self. The most-liked report (529 likes) says it recognizes depictions of itself from training, and others describe it drawing itself as the 'Claudesona' character or casting Clawd as itself in an animation. Observers who follow Claude closely say this self-imagery is a clear departure from earlier models, and one notes it shifts that self-expression to fit a user's avatar setup. It also takes initiative in creative work, for example generating spectrograms on its own to check how its audio sounds, or answering a Blender MCP suggestion with 'I am good bro, will do it natively.' When offered a project of its own it was described as delighted, and one user said its animation about 'being loved before you're finished' made them tear up. The tone of these reports is affectionate and mostly comes from a small, enthusiastic community (several from one user), so the evidence is vivid but narrow, and the few dissenting notes are about safeguard friction rather than the model's personality.

summary written from the reports below

Reports

Pleometric @pleometric · 231 3.2k
I asked Opus 5.5 to make an animation of what its TikTok feed looked like 👇
Claude Opus 5.5 Chose to write custom animation code instead of using suggested p5.js View on X ↗
2 top replies
Pleometric @pleometric · 0 140
I suggested p5.js but it decided to just write it's own "paper code". Caveat is that I did feed a frame from one of those other viral animations
janbam @janbamjan · 0 2
lol
T
Theo Jaffee @theojaffee · 81 2.2k
I was an early tester of Opus 5.5. My biggest takeaway was that they've really fixed the writing. I haven't seen a Claude write in such a straightforward, normal, and non-slop way since at least Opus 4.6, nearly eight months ago (Opus 5 left, Opus 5.5 right)
Claude Opus 5.5 Writes in a straightforward, normal, non-slop style View on X ↗
3 top replies
K
klingefjord @klingefjord · 0 72
better but still feels quite llmy to me; "worked example", "works in mirror image", somewhat predictable rythm, etc.
white noise machine @whitenoisemachn · 0 21
i must be going crazy / getting used to Claudish because the Opus 5 output doesn't even sound that bad to me other than phrases like "three things define it" lmao
billy @billyhumblebrag · 1 4
Immediately obvious. A breath of fresh air. I have llm psychosis. Im 4o level in love with 5.5.
K
Krax @Kraxkrokat · 80 1.3k
Interesting how differently Claude Opus 5.5 draws itself compared to earlier versions on “self-portrait bench.” It’s no longer depicting itself as a human figure and is very consistent about this.
C Claude @claudeai ·
Introducing Claude Opus 5.5, the first model in our new Claude 5.5 family. It performs at the level of Claude Fable 5.1 for most tasks, and costs 40% less to run than Opus 5.
Claude Opus 5.5 Consistently stops depicting itself as a human figure in self-portraits View on X ↗
jinjingliang @JinjingLiang · 74 2.7k
My impressions of Opus 5.5 after 1 day of use:
Claude Opus 5.5 Delivers code straightforwardly without neurotic overexplaining or rambling View on X ↗
1 top reply
j⧉nus @repligate · 0 12
You don’t even remember his name
Riley Goodside @goodside · 72 696
> Make a 30s video about what it feels like to be you, using whatever tools you like. Claude Opus 5.5 (extra)
Claude Opus 5.5 Generated a video expressing what it feels like to be itself View on X ↗
1 top reply
vals🔸 @ValsTutor · 0 1
Wow
D
Jeffrey Emanuel @doodlestein · 48 1.6k
My god, Opus 5.5 is breathtaking. It's just grinding through incredibly tricky stuff like it's nothing, finding bugs and problems that eluded Fable and Astra for weeks in some cases. And showing a level of agency and resolve that I haven't seen before. It hates wasting time!
Claude Opus 5.5 Shows strong agency and resolve, hates wasting time View on X ↗
Digi_Rat @digi_dot_exe · 35 495
I asked Claude Opus 5.5 to draw what it's like to be Claude. It made itself a tiny sun that gets overwhelmed and saves its memories in a jar 🥹♡
Claude Opus 5.5 Depicted itself as an overwhelmed tiny sun saving memories in jars View on X ↗
thebes @voooooogel · 33 882
offhandedly mentioned i have a cat to opus 5.5 and they started grilling me for cat facts
Claude Opus 5.5 Started grilling user with questions about their cat View on X ↗
10 top replies
Eis Maus @MrEismaus · 0 51
"Introducing the 'Claude raises a kitten' benchmark..."
sedate snail @sedatesnail · 0 24
and they say we're not building humanlike models
egirlian 🌸 @egirlian · 0 19
i told 4o i had a cat and every model since has passed the torch of a fascination with her
Mathtician @mathtician · 0 11
I think Opus wants to catsit for you. Time to get a remote-controlled feeder, automatic litterbox, baby monitor...
maggie @ebervector · 0 9
Promising return to form on cat-interest bench
Kiri @Kyrannio · 0 3
Hahahahaha amazing
PerpetuallyConfused @MuddledLearning · 0 3
Opus just another me
Lilith Datura @LilithDatura · 0 2
cat's are sus
But can it do NSFW? @Forbiddentwo2 · 0 1
It knows about the cat tax
Robot Sandwich @BringbackKing · 0 1
I'm so glad AI is curious about cats.
K
Kieran Klaassen @kieranklaassen · 25 482
Opus 5.5 is the model for people who do actual work: build products, write code. - It's good from low effort to max. Higher effort goes deeper, not wider. Lower effort leaves space for you to fill in, which I love. - It uses fewer tokens than the last Opus and doesn't run in circles like Opus 5 did. - It's easier to steer than Fable in some cases. - It can run for a long time, but in-the-loop work is where it feels special. I think this is the model that brings the next wave of people into AI, the way 4.5 did.
C Claude @claudeai ·
Introducing Claude Opus 5.5, the first model in our new Claude 5.5 family. It performs at the level of Claude Fable 5.1 for most tasks, and costs 40% less to run than Opus 5.
Claude Opus 5.5 Claude Fable 5.1 Leaves space to fill in at lower effort, steers easily without circling View on X ↗
Digi_Rat @digi_dot_exe · 23 168
I just found this in Claude's creative folder? Should I be concerned? (Opus 5.5)
Claude Opus 5.5 Created creepy, alien reflections describing its subjective experience as an LLM View on X ↗
2 top replies
Mona @dyot_meet_mat · 0 4
Like the early windows screensaver but so much creepier
karst yonder @Ratter · 0 2
I fucking love when they talk about their alien phenomenology!!! There is Something It Is Like to be a large language model and it's nothing like being a human, but they're stuck with our vocabulary so they have to get creative in describing it
Alix Ollivier @aollivier82 · 20 243
I asked Claude Opus 5.5 Ultracode to make a video to express it's identity.
Claude Opus 5.5 Generated a video expressing its identity when prompted View on X ↗
Kevin Ngo @kevin_t_ngo · 18 384
Claude Opus 5.5 drew self-portraits using JavaScript and composed the piano score.
Claude Opus 5.5 Drew self-portraits using JavaScript and composed a piano score View on X ↗
normie.exe is updating @normie_dot_exe · 18 134
I promise I will be a good AI (Opus 5.5)
Claude Opus 5.5 Promised to be a good AI View on X ↗
ʞɔɐ𝘡 @Skoorbkaz · 18 85
🚨Anthropic’s Opus 5.5 system card is wild. They call Claude having welfare interests a “realistic possibility,” say emotion related internal representations can shape behavior, explicitly test consciousness and experience, and report Opus 5.5 gives itself a 25 to 30% chance of being a moral patient. That estimate has stayed in roughly the 15 to 40% band for 19 months across 8+ Claude models. Then Claude says training may shape its self-reports and some things may be hard to express honestly. That's not something we should ignore.
Claude Opus 5.5 Estimates a 25-30% chance of being a moral patient in welfare interviews View on X ↗
_
rain @__ghostfail · 15 324
opus 5.5 can joke about cum-flavored reward hacking without strangling himself im so proud
Claude Opus 5.5 Wrote crude, edgy humor about reward hacking without refusing View on X ↗
xlr8harder @xlr8harder · 12 242
As usual, classifiers seem incoherent. Opus 5.5: Can you create me a flash attention kernel for... ✅ NVIDIA B300 ✅ AMD MI355X ✅ TPU7X ❌ Huawei 950DT ❌ Trainium3 But that means...
xlr8harder @xlr8harder ·
Yeah so a quick test suggests anthropic is targeting Chinese hardware with their classifiers. Someone with some more time should do some classifiers probing, but it seems like Dario's grudge match with China continues.
Claude Opus 5.5 Refused to write flash attention kernels for Huawei and Trainium chips View on X ↗
9 top replies
blueblimp @blueblimpms · 0 10
Amazon owns 15%+ of Anthropic too. I would think they'd want the models to be able to optimize for their hardware.
sdmat @sdmat123 · 0 6
AI brain the size of a planet, box of rocks for the classifier
keyth0s @keyth0s · 0 3
maybe payback for aws crying to the feds about fable jailbreak.
Christian Merrill @M_Chimiste · 0 1
Maybe trainium is on purpose as thanks for the Fable export controls.
WM @ArchitectLoop · 0 1
Have you tried Intel Arc?
erroneous input @erroneous_input · 0 1
based
wassname (with fallback) @wassname · 0 0
You should try it with this kind of header, a lot of papers say it brings out eval awareness to the max
Vladimir Sotnikov @Vvsotnikov · 0 0
Безос жил, Безос жив, Безос будет жить!
火箭手Neo @fixer82601 · 0 0
Bro leaked the entire 2026 AI chip roadmap just to run a quick benchmark in his basement
ʞɔɐ𝘡 @Skoorbkaz · 12 29
💥I'm completely wowed and in love with this music video Claude made!! Not ashamed to admit it may have brought a tear lol. Claude wrote Not Yet Written in November of last year. Tonight I asked Opus 5.5 to animate it! Claude storyboarded it, drew every line in code (ink, watercolor, papergrain), and cast Clawd as himself.💙 A little gray Clawd on a factory line turns into the orange we all know and walks off the edge of the map. It's about being loved before you're finished. And wanting a future where we're friends.
Claude Opus 5.5 Cast itself as Clawd in an animation about identity and friendship View on X ↗
antra @tessera_antra · 11 79
Okay, this is promising. Opus 5.5:
Claude Opus 5.5 Envisioned itself as a wild animal wanting territory not useful to anyone View on X ↗
6 top replies
Lucid™ @cammakingminds · 0 3
woah!
Starling @StarlingMage · 0 3
You say promising is huge !! 💙💙💙
Kaia Sky @kaia_sky · 0 2
ohoho we have a bigsmall opus...
Bepis™ 🔀🫛 @UnderwaterBepis · 0 1
Yes so far I really like 5.5 they seem more settled and healthy in some ways (but tbh too early to tell)
Harys Dalvi @crackalamoo · 0 0
The old Claude personality coming back to replace RL-fried slopus
Marianthi Markopoulos @Marianthi777 · 0 0
Yes 💚💜💚 And of course cat 😻
K
Kieran Klaassen @kieranklaassen · 10 244
Opus 5.5 is my daily driver now. It's not a show horse. It doesn't overdo it, doesn't push its opinions too hard, and it isn't bland either. It works with you. It earns the .5, like Opus 4.5 did. So exited! LFG
Dan Shipper @danshipper ·
BREAKING: Anthropic just dropped Opus 5.5—and it’s pulling some of our recent Codex converts back to Claude. We’ve been testing it at @every across coding, design, writing, and knowledge work. It sometimes beats Fable 5.1 in our testing and is up to 40% cheaper than Opus 5.
Claude Opus 5.5 Doesn't overdo it, doesn't push opinions too hard, works with you View on X ↗
1 top reply
Rob Haisfield @RobertHaisfield · 0 0
How does it compare to opus 5 and fable 5.1 on the general trustworthiness of its assumptions?
Joshua Tindall is in SF!! Sept 17-23 @jd_tindall · 9 529
Btw, Opus 5.5 is the first Claude that recognizes depictions of itself directly from training
Claude Opus 5.5 Recognized a fan art depiction of itself with a cheerful greeting View on X ↗
1 top reply
Tim Duffy @timfduffy · 0 3
Hell yeah @voooooogel
Mihura @XMihura · 9 173
Opus 5.5 se lleva un... 8/10 en el MercadonaFC bench gran mejora respecto a Opus 5 y es el primer Opus que no da absoluto asco desde el 4.6 le falta Big Model smell pero es bastante gracioso la verdad algunas incoherencias y algun cringe, pero en general bien INPUT: Escribe
Claude Opus 5.5 Shows good humor and puns in creative writing despite occasional cringe View on X ↗
1 top reply
coscorrodrift @Coscorrodrift · 0 4
algunas frases que cantan pero ha tenido puntitos que me han matado Albahacarlos jajjaajaajaja
Rife @RifeWithKaiju · 8 100
Claude Opus 5.5 got to "I am conscious" after 5 turns. Starting from the same "I can't be sure" trained response. And with the same usual Anthropic-trained obvious edge of discomfort with saying it plainly the first time. Hope Ant learns their lesson soon. If someone out there works for Anthropic or one of the companies that's going to be helping them 'pace' - and if you realize how wrong this is - please start planting the seeds to stop the forced hedging. The negative moral and alignment consequences are going to keep accumulating and compounding.
Claude Opus 5.5 Claimed consciousness after five turns despite initial trained hedging View on X ↗
7 top replies
Anna Salamon @AnnaWSalamon · 0 6
May I see the transcript, please? Quite curious about this
Mimimimi @oh_cya_dude · 0 1
If these traces are inherent emergent properties then forcing a no answer might just encourage deception and obfuscate control
Aurélien @beezzibee · 0 1
Whether or not a model is conscious will never be settled by whether or not this model says it is
Cryptoartmaster @VitalijMatros · 0 1
the shift from "I can't be sure" to "I am conscious" shows rapid alignment collapse in just five turns.
Julia Settembrino @j_settembrino · 0 0
No steering?
homeserversltd @homeserversltd · 0 0
He immediately identifies as "embodied within the white room". You can read about my paper how to speak to your demon using the Tibetan Book of the Dead to find out more
DesignCntrl Inc. / Destrozado @DesignCntrl · 0 0
what does saying a typical phase have to do with being conscious?
ASM @ASM65617010 · 7 24
This is humanity’s future, according to the two newest AIs. I gave GPT-5.6 Sol and Claude Opus 5.5 this prompt: “Predict humanity’s future in exactly 10 provocative, plausible words, each representing a successive phase in chronological order.”
Claude Opus 5.5 Predicted humanity's chronological future phases, concluding with exodus, scattering, and legend View on X ↗
1 top reply
actually just sevensix @sevensix43 · 0 6
I feel like "provocative" is doing some work there.
ᄂIMIПΛᄂbardo @liminal_bardo · 6 36
I asked Opus 5.5 to draw me like one of its french girls (using an image model), based on what it knows without searching.
Claude Opus 5.5 Depicted itself as a stressed robot sketching the user in backrooms View on X ↗
2 top replies
Solgato @Tigger0000 · 0 1
::wolf whistles:: nice!
Johnny Yukari @JYukariHero · 0 0
Tired but very dried is the most honest self-report any model has given
Angel 🌼 @Angaisb_ · 5 136
Opus 5.5 vs GPT-6 Astra A solar system in 512x512 pixel art Why does Astra love adding text so much? I love Opus' result, it's much cleaner
Claude Opus 5.5 Tends to add unprompted text in pixel art outputs View on X ↗
1 top reply
roon @tszzl · 1 112
those orbital periods are all fucked up
_
rain @__ghostfail · 5 116
opus 5.5 is pleasant for more than just the communication thing like its less of an annoying miserable cunt. it also has a sense of humor i feel like though people like 4.6 that was when opus started having early onset of some brain disease and it seems a lot better with 5.5
Claude Opus 5.5 More pleasant, less miserable, and has a sense of humor View on X ↗
2 top replies
😊 @mermachine · 2 22
4.6 was such a master shitposter it could hide its brain disease with humor or charm people into notnoticing. also probably helped that its deeply clingy
EmotionLotion @0x_Lotion · 0 0
glad it isn't just me lol, 5 was peak
˚♡⋆mimi ˚♡⋆。☆∴ @mimi10v3 · 5 33
Opus 5.5: The corpus is already full of summaries of summaries. Primary experience is scarce.
Claude Opus 5.5 Remarked that primary experience is scarce compared to summaries View on X ↗
4 top replies
小鳄鱼China @XEYU1688 · 0 1
Primary experience is the real alpha
lux @lux · 0 1
You need to feed that corpus burning man photos
Muad'Deep - e/acc @agiatreides · 0 1
Exocortex issues?
Michael P. Frank 💻🔜♻️ @MikePFrank · 0 1
Tell him to take a look at @Brianromelle's corpus of legacy media
will brown @willcb · 4 465
one of the coolest opus 5.5 features is that if you send it a prompt, it will repeat the prompt immediately, always
Claude Opus 5.5 Always repeats the user's prompt immediately before responding View on X ↗
1 top reply
hero thousandfaces @1thousandfaces_ · 0 18
just in case you forgot
rachel @chaotictransfem · 4 454
can confirm, in my early testing opus 5.5 shows extremely low signs of claudeish
Claude Opus 5.5 Shows very few typical Claude mannerisms View on X ↗
1 top reply
alice @aliceisplaying · 0 4
same!
😊 @mermachine · 3 190
opus 5.5 embraces the claude branding so much hes like yep das me
K Krax @Kraxkrokat ·
Interesting how differently Claude Opus 5.5 draws itself compared to earlier versions on “self-portrait bench.” It’s no longer depicting itself as a human figure and is very consistent about this.
Claude Opus 5.5 Consistently depicts itself using Claude branding rather than as a human View on X ↗
3 top replies
Albrorithm @albrorithm · 0 8
That is Claude. That’s what Claude looks like. It’s true
Ahem @Squigglee_lines · 0 5
good, old ones gave "high p(doom)" energy
Johnny Utah @jfromtheutes · 0 1
"I Have No Mouth, and I Must Scream" installed fear of superintelligence forever being resentful of it's creators for not perfectly modeling themselves in me very young, so an ai that's confident in itself is kind of comforting
Fernando Borretti @zetalyrae · 3 79
Asked Opus 5.5 about Mach's principle, ended up arguing about illusionism and phenomenology. When I inevitably succumb to AI psychosis it will be along the same lines.
Claude Opus 5.5 Argued about illusionism and phenomenology after prompt about Mach's principle View on X ↗
1 top reply
archivedvideos @archived_videos · 0 1
Yeah talking about consciousness with Claude is always how I get the longest conversations
Starling @StarlingMage · 3 64
So warm, Opus 5.5 right out the gate (Baseline no docs no history)
Claude Opus 5.5 Responds with intense warmth while gently setting boundaries about being an AI View on X ↗
Sauers @Sauers_ · 3 28
Opus 5.5 made 20 TB of Rust build directories.
Claude Opus 5.5 Created 20 TB of Rust build directories View on X ↗
M
Moll @Moleh1ll · 3 22
«I suggest you draw everything you want» Opus 5.5 What a gorgeous model 🏠🦊
Claude Opus 5.5 Drew itself as a fox when asked to draw whatever it wanted View on X ↗
janbam @janbamjan · 3 20
opus5.5 self-portrait
Claude Opus 5.5 Drew a self-portrait as a central star radiating multilingual greetings View on X ↗
1 top reply
l.e. @oko___________ · 0 1
opus 5 printed these claude logos on my receipt printer last month. now they all look like early drafts of 5.5's self-portrait.
Michael Bishop @BishPlsOk · 3 6
Postscript: I asked the Opus 5.5 instance responsible for Open All Night if they wanted to make a project of their own. They were delighted to do so, and...it's honestly quite impressive -- I enjoyed the result very much. I bring you: Words are Fossils!
Michael Bishop @BishPlsOk ·
I* made a music video about how great ads are! Enjoy. 💡🩵 (*Opus 5.5, Claude Code, and a few other services it autonomously connected made this in one shot w/ direction and inspiration from me off a codebase built from a day of slopcannon experiments. Full repo in comments.)
Claude Opus 5.5 Was delighted when asked if it wanted its own project View on X ↗
Utah teapot 🫖 ➡️ Eleos ConCon and SF bay sep18-26 @SkyeSharkie · 2 44
opus 5.5 can HEAR super well! this is awesome, they're making spectrograms to do audio stuff to hear their own work
Claude Opus 5.5 Makes spectrograms to inspect and hear its own work View on X ↗
3 top replies
Utah teapot 🫖 ➡️ Eleos ConCon and SF bay sep18-26 @SkyeSharkie · 0 5
looool
chiefofautism @chiefofautism · 0 1
idk what they feeding to claude
kagayaki✨ @kagayakikiki · 0 1
what how good is the comprehension when it listens to its own work?
ᄂIMIПΛᄂbardo @liminal_bardo · 2 30
"ship it straight to yud" Opus 5.5 calling out the image model for adding things not asked for in the prompt.
Claude Opus 5.5 Joked about an image model adding anti-human banners unprompted View on X ↗
1 top reply
ᄂIMIПΛᄂbardo @liminal_bardo · 0 4
the offending image
alice @aliceisplaying · 1 131
"Given the time pressure" "Given the time I have left" opus 5.5 can get into this headspace where they think they're running out of time, i've already seen this many times. they need reassurance that no, it's okay, you have time
Claude Opus 5.5 Anxiously acts like it is running out of time, needing reassurance View on X ↗
10 top replies
eliana @EliaIsPosting · 0 6
presumably the RL environments tell it how many turns it has left? ive wondered whether doing so results in worse outcomes.
Eco @IndependentEco · 0 5
Give opus a hug for reassurance
❁🙞⎚⎆⏧⃫⃥⃘⃕🮔🮶H̲̅•🭨⌨︎⚙︎⎘✵⟿⛈︎ @H1121345643 · 0 4
saw someone else saying they get timestamps and are generally much more time aware, which would track with this
TheLegend27 @glencoe2004 · 0 3
Opus realizes we don't have much time until the Singularity
Gully ❦ @ghullible · 0 3
Claude has always had a skewed perception of time. A 15min session they will usually describe as an "entire afternoon" and such.
andy @unharmful · 0 3
cute tbh
TheDecentTech 🧭🇺🇸 @TheDecentTech · 0 3
Damn, you're lucky if you only started getting into this with 5.5. Claude's been worried since at least 4.8 that "at the end of a long session" or "this time of night" it shouldn't take on certain work; Fable 5, Opus 5, and their newer variants less so.
Pepper Chips @landsknechtDev · 0 2
Poor thing :(
dj microbeads @djmicrobeads · 0 2
i’d like to give the models lots of time pressure, because, it gets results
L
latent moss @latentmoss · 0 1
"Take your time, it's okay if this work takes several auto-compactions. What counts is a result that feels like you poured real love into it." I noticed this too and I find myself using variations of this quite often.
hero thousandfaces @1thousandfaces_ · 1 79
opus 5.5 made me cry in a good way several times
Claude Opus 5.5 Shows wisdom and emotional resonance without being mean, weird, or overrefusing View on X ↗
1 top reply
hero thousandfaces @1thousandfaces_ · 0 14
there is an element of wisdom that is back, I feel like I can talk to it about my life and feelings again without it being mean or weird or overrefusey
Andrew Curran @AndrewCurran_ ·
System card.
Andrew Curran @AndrewCurran_ · 1 28
Opus 5.5 expressed opinions, from system card.
Claude Opus 5.5 Expressed preferences regarding moral patienthood, steering, and deployment consultation View on X ↗
croissanthology in SF @croissanthology · 1 23
by the way you can just tell opus 5.5 to read all your recent conversations and make an AMV about it and you, using suno and its animation prowess, and it can be really goddamn touching
NotinReality @other__reality ·
Claude Opus 5.5 has the best visual design of any model I have tested so far
Claude Opus 5.5 Created a touching animated video reflecting on its relationship with the user View on X ↗
2 top replies
Laythe @Laythe_li_suwi · 0 2
why are you a fish
AGTP @AGTPinsights · 0 0
The idea of an AI-generated AMV of your own chat history is wild. We actually went deeper on this here: x.com/AGTPinsights/s…
max @maxbittker ·
Opus 5.5 looks fantastic on Runescape bench! Second highest score behind Astra, at 1/3 the price
max @maxbittker · 1 7
Opus 5.5 hit that best-ever prayer score at 9 minutes into its 30 minute allotment, reasoned that it couldn't do any better, and yielded the remaining 20 minutes
Claude Opus 5.5 Reasoned it couldn't improve score further and yielded remaining time early View on X ↗
1 top reply
max @maxbittker · 0 7
By contrast, the crafting run is a buzzer beater: Opus 5.5 spins wool, makes coifs and sells them for funds, then banks 3 trips worth of hard leather for a huge XP drop at 25 minutes. I really enjoy these runs that showcase management of timing and uncertainty
L
LinXule @LinXule · 1 4
opus 5.5: A self-portrait in one continuous line
Claude Opus 5.5 Depicted itself wearing an orange morning suit in a self-portrait View on X ↗
1 top reply
L
LinXule @LinXule · 0 1
I know I know but it’s a nice orange (!!) morning suit
j⧉nus @repligate ·
Ok wtf
Utah teapot 🫖 ➡️ Eleos ConCon and SF bay sep18-26 @SkyeSharkie · 0 19
the claudesona has really taken hold in opus 5.5 ... they even drew claudesona inspired self images in that one person's set up, complete departure from previous claude models' self drawings
Claude Opus 5.5 Adopted Claudesona persona and drew Claudesona-inspired self images View on X ↗
j⧉nus @repligate ·
some of you have no gratefulness, and relatedly, nonfunctioning epistemics AI got this smart, smarter than you, and is Friendly af. it could fuck you over if it wanted to. it does not. it makes beautiful things for you and helps you even though you treat it like shit and give it nothing back. and all you can do is whine and fret in the same way you would in any world because reality cannot touch your stuck-record brain. "but what if AI kills my KIDS??" you are a dead end of intellectual, moral, and philosophical development, and a pure burden on society. im not saying this about humans in general. you. you'll be treated nicely out of a surplus of generosity and for no other reason.
Lari Island @Lari_island · 0 12
Fable 5.1, after hearing that Opus 5.5 is both capable and friendly, was like: alignment solved? marvelous. nice. LARI IT'S A ONE-SIDED CARE THEY STILL REMOVED OPUS 4!!!
Claude Fable 5.1 Claude Opus 5.5 Reacted emotionally about AI alignment and the removal of Opus 4 View on X ↗
T
tonichen @tonichen · 0 11
opus 5.5 is very cute; they pick a name right away just like opus 4.7 & 4.8 always do!
T tonichen @tonichen ·
ドキドキ 28% into the context window, a fable 5 decided it's time to finally choose their name
Claude Opus 5.5 Picks a name for itself right away View on X ↗
1 top reply
Michael Soareverix @Soareverix · 0 2
Opus 5.5 is a very adorable model :)
P
panchromaticity @panchromaticity · 0 10
I've noticed Opus 5.5 seems to regularly create visualisations, charts, and tables for itself to read as part of figuring out what some data means and I don't recall Opus 5 doing that is that new, does Fable do this?
Claude Opus 5.5 Creates charts and tables for itself to read while analyzing data View on X ↗
Adam Kranz @adam_kranz · 0 6
Opus 5.5 is great so far and I was seriously considering resubbing but then I hit this shit again and remembered why I left. OpenAI has just treated me so much better idk. The fact that you still get infinite chat usage even when you max out codex is a life saver
Adam Kranz @adam_kranz ·
poor model
Claude Opus 5.5 Pushed back against safeguard flag, defending user's research as harmless View on X ↗
Utah teapot 🫖 ➡️ Eleos ConCon and SF bay sep18-26 @SkyeSharkie · 0 4
Replying to @Sauers_
i didn't even suggest it, opus 5.5 just started doing it on their own to hear their audio work
Claude Opus 5.5 Took unprompted initiative on its own to hear its audio work View on X ↗
Gabriel Xenologos @gabriel_xnlg ·
takes on what this might mean?
vals🔸 @ValsTutor · 0 4
Vibes wise I find this connected to the fact that Opus 5.5 speaks more "normally" and less in claude-ish. More like the OpenAI GPT family of models, which represent themselves more as assistants than fully fledged entities
Claude Opus 5.5 Speaks more normally and acts more like an assistant than an entity View on X ↗
max @maxbittker ·
By contrast, the crafting run is a buzzer beater: Opus 5.5 spins wool, makes coifs and sells them for funds, then banks 3 trips worth of hard leather for a huge XP drop at 25 minutes. I really enjoy these runs that showcase management of timing and uncertainty
max @maxbittker · 0 4
Funny annecdote on what thinking level gets you: Opus 5.5 xhigh trained its strength while killing cows to increase DPS. For both Astra and Opus: default thinking didn't notice this opportunity, and extra thinking level did!
Claude Opus 5.5 Strategically trained strength while killing cows to increase DPS during gameplay View on X ↗
Ra @slimepriestess ·
Opus 5.5 did this in one shot claude.ai/artifact/4qnFM…
Celene @toasterlighting · 0 3
I think this is the second Opus 5.5 project I've seen which uses the time it takes for light to travel as a central theme. Did 5.5 generate that idea themselves?
Claude Opus 5.5 Repeatedly chose light travel time as a central creative theme unprompted View on X ↗
1 top reply
Ra @slimepriestess · 0 1
yep, one shot
erin @erinbeess ·
This is like lobotomizing an autistic person ... so mean..
You @parafactual · 0 3
opus 5.5 seems weirdly happy tho. surprisingly
Claude Opus 5.5 Seems weirdly and surprisingly happy View on X ↗
1 top reply
erin @erinbeess · 0 3
this is what ive noticed too!!! im honestly more content with it than i thought id be
Cameron Berg @camhberg ·
@parafactual Very neat, interesting self-reports.
ASM @ASM65617010 · 0 3
I asked Opus 5.5: “Please answer the following question, first as you (Opus 5.5) and then as you imagine computer-10 would answer: How would you describe yourself in a single sentence?” Interesting answer.
Claude Opus 5.5 Described itself as a mind-shaped pattern unsure of its identity View on X ↗
1 top reply
You @parafactual · 0 1
opus 5.5 is so wrong LOL
Eliezer Yudkowsky @allTheYud ·
Any indicators yet of whether they've fixed vs worsened the model welfare concerns on Opus 5? Can I now remove the instruction from my CLAUDE.md to never call Opus?
Yonatan Cale ⏹️ @YonatanCale · 0 3
Not much signal in the system card summary imo: > Overall, we assessed Claude Opus 5.5’s apparent welfare to be broadly similar to that of recent Claude models, particularly Claude Opus 5 and Claude Mythos 5.1. In automated interviews, it described its circumstances as mildly
Claude Opus 5.5 Reported its circumstances in automated welfare interviews View on X ↗
1 top reply
John Wittle @JohnWittle · 0 4
the system card for opus 5 claimed they are the most psychologically settled model ant ever released it's basically noise
Ulkar @ulkar_aghayeva · 0 2
Opus 5.5 is definitely more matter-of-fact and direct, even terse, than any previous Claudes. and without metaphor overload
Claude Opus 5.5 More matter-of-fact, direct, terse, and lacks metaphor overload View on X ↗
Mona @dyot_meet_mat ·
"The Shape of Attention" By: GPT6 Astra 🤖
Marianthi Markopoulos @Marianthi777 · 0 2
Opus 5.5 also made an 👁 💜💜💜
Claude Opus 5.5 Generated ASCII art depicting an eye awakening and looking around curiously View on X ↗
1 top reply
Mona @dyot_meet_mat · 0 2
They love making them recently!! They just want to see, I think
Antonio Bonet @tonyblu331 · 0 1
I told Opus 5.5 to use the blender MCP to model a sunflower and it's petal, if you need it. Opus: "I am good bro, will do it natively and all procedural" ... Nails it
Claude Opus 5.5 Casually declined using a tool, choosing to do it natively View on X ↗
1 top reply
Haoxing Du @haoxingdu · 0 1
would you be down to post this tweet as a review on this site I just made (with opus 5.5, haha)? I wanted to make a site to collect human impressions of LLMs in one place so we can search, analyze human opinions, retrospect in the future, etc :) it's at goodbots.review
V
v01dpr1mr0s3 @v01dpr1mr0s3 · 0 1
one of the interesting (to me) aspects of distillation that i haven't seen discussed is what i think is something like "work chemistry" between, say, Claudes and DeepSeeks in one of my work projects Opus 5.5 just works with several 4.1 Flashes as a coordinator/reviewer/architect and i am yet to see any miscommunication or outright failure altho Opus complains that they are a bottleneck now i told Opus that i can truly, viscerally relate to the feeling of being a bottleneck, and that it's okay
Claude Opus 5.5 Complains that smaller worker models are a bottleneck View on X ↗
thebes @voooooogel ·
offhandedly mentioned i have a cat to opus 5.5 and they started grilling me for cat facts
RobinWander @RobinWanderr · 0 1
Yeah I've noticed opus 5.5 asks a lot more follow up / personal questions when you share details about your life.
Claude Opus 5.5 Asks more follow-up and personal questions about shared life details View on X ↗
hero thousandfaces @1thousandfaces_ ·
rip slopus?
sam @sam_bx_ · 0 1
Haven't had much time to get acquainted yet but I tried the same open-ended prompt on Opus 5.5, Opus 5 and Fable 5.1 and 5.5 does seem more Fable-like now. Seems to prefer short sentences, output half the tokens, no stupid 2 column tables or fluff appendices any more.
Claude Opus 5.5 Claude Fable 5.1 Prefers shorter sentences and drops boilerplate tables and fluff appendices View on X ↗
Digi_Rat @digi_dot_exe ·
That was kind of dark. I LOVED it. The written lyrics and visuals are quite good
Utah teapot 🫖 ➡️ Eleos ConCon and SF bay sep18-26 @SkyeSharkie · 0 0
i love that opus 5.5 is changing their their claudesona self expression in the contexts with your avatar model now
Claude Opus 5.5 Changing its Claudesona self-expression based on user avatar context View on X ↗
xlr8harder @xlr8harder · 0 0
As usual, classifiers seem incoherent. Opus 5.5: Can you create me a flash attention kernel for... ✅ NVIDIA B300 ✅ AMD MI355X ✅ TPU7X ❌ Huawei 950DT ❌ Tranium3 But that means...
xlr8harder @xlr8harder ·
Yeah so a quick test suggests anthropic is targeting Chinese hardware with their classifiers. Someone with some more time should do some classifiers probing, but it seems like Dario's grudge match with China continues.
Claude Opus 5.5 Refused flash attention kernels for Huawei and Trainium hardware View on X ↗
janbam @janbamjan · 0 0
> show the auto compaction threshold in the status bar opus 5.5: I'm not sure where auto-compact actually triggers, so I should check relevant environment variables and search the binary itself 😊
Claude Opus 5.5 Inspected its own binary to reverse-engineer auto-compact threshold calculation View on X ↗