Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Claude Opus 4.5

anthropic.com

Post reply on HN
Print view View original

431–440 of 525 posts

Re: Claude Opus 4.5

#431

Post by jaggs »

riwsky wrote: post #410
blurbleblurble wrote: post #228

Earlier quoted context omitted.

I'm working on a hard problem recently and have been keeping my "model" setting pegged to "high". Why in the world, if I'm paying the loss leader price for "unlimited" usage of these models, would any of these companies literally respect my preference to have unfettered access to the most expensive inference? Especially when one of the hallmark features of GPT-5 was a fancy router system that decides automatically wh…

Because intentionally fucking over their customers would be an impossible secret to keep, and when it inevitably leaks would trigger severe backlash, if not investigations for fraud. The game theoretic model you’re positing only really makes sense if there’s only one iteration of the game, which isn’t the case.

That is unfortunately not true. It's pretty easy to mess with your customers when your whole product is as opaque as LLMs. I mean they don't even understand how they work internally.

Re: Claude Opus 4.5

#432

Post by dave1010uk »

aurareturn wrote: post #418
dave1010uk wrote: post #249

The Claude Opus 4.5 system card [0] is much more revealing than the marketing blog post. It's a 150 page PDF, with all sorts of info, not just the usual benchmarks. There's a big section on deception. One example is Opus is fed news about Anthropic's safety team being disbanded but then hides that info from the user. The risks are a bit scary, especially around CBRNs. Opus is still only ASL-3 (systems that substantia…

Pages 22–24 of Opus’s system card provide some evidence for this. Anthropic run a multi-agent search benchmark where Opus acts as an orchestrator and Haiku/Sonnet/Opus act as sub-agents with search access. Using cheap Haiku sub-agents gives a ~12-point boost over Opus alone. Will this lead to another exponential in capabilities and token increase in the same order as thinking models?

Perhaps. Though if that were feasible, I'd expect it would have been exploited already.

I think this is more about the cost and time saving of being able to use cheaper models. Sub-agents are effectively the same as parallelization and temporary context compaction. (The same as with human teams, delegation and organisational structures.)

We're starting to see benchmarks include stats of low/medium/high reasoning effort and how newer models can match or beat older ones with fewer reasoning tokens. What would be interesting is seeing more benchmarks for different sub-agent reasoning combinations too. Eg does Claude perform better when Opus can use 10,000 tokens of Sonnet or 100,000 tokens of Haiku? What's the best agent response you can get for $1?

Where I think we might see gains in _some_ types of tasks is with vast quantities of tiny models. I.e many LLMs that are under 4B parameters used as sub-agents. I wonder what GPT-5.1 Pro would be like if it could orchestrate 1000 drone-like workers.

Re: Claude Opus 4.5

#433

Post by mannanj »

marcus_holmes wrote: post #341
mannanj wrote: post #329

Earlier quoted context omitted.

I once heard a devils advocate say, “if child porn can be fully AI generated and not imply more exploitation of real children, and it’s still banned then it’s about control not harm.” Attack away or downvote my logic.

I think this is a serious question that needs serious thought. It could be viewed as criminalising behaviour that we find unacceptable, even if it harms no-one and is done in private. Where does that stop? Of course this assumes we can definitely, 100%, tell AI-generated CSAM from real CSAM. This may not be true, or true for very long.

If AI is trending towards being better than humans at intelligence and content generation, it's possible its CGP (Child generated P*n) would be better too. Maybe that destroys the economies of p*n generation such that like software generation, it pushes people away from the profession.

Re: Claude Opus 4.5

#434

Post by quleap »

roywiggins wrote: post #339
zsoltkacsandi wrote: post #197

Earlier quoted context omitted.

> The nerf is psychologial, not actual Once I tested this, I gave the same task for a model after the release and a couple weeks later. In the first attempt it produced a well-written code that worked beautifully, I started to worry about the jobs of the software engineers. Second attempt was a nightmare, like a butcher acting as a junior developer performing a surgery on a horse. Is this empirical evidence? And this…

https://en.wikipedia.org/wiki/Regression_toward_the_mean The way this works is: 1) x% of users have an exceptional first experience by chance. Nobody who has a meh first experience bothers to try a second time. 2) x²% of users also have an exceptional second experience by chance 3) So a lot of people with a great first experience think the model started off great and got suddenly worse Suppose it's 25% that have a re…

Your theory does not hold if a user initially had great experience for weeks and then had bad experience also for weeks.

Re: Claude Opus 4.5

#435

Post by fakwandi_priv »

blurbleblurble wrote: post #192
lukev wrote: post #154

Earlier quoted context omitted.

There are two possible explanations for this behavior: the model nerf is real, or there's a perceptual/psychological shift. However, benchmarks exist. And I haven't seen any empirical evidence that the performance of a given model version grows worse over time on benchmarks (in general.) Therefore, some combination of two things are true: 1. The nerf is psychologial, not actual. 2. The nerf is real but in a way that…

I'm pretty sure this isn't happening with the API versions as much as with the "pro plan" (loss leader priced) routers. I imagine that there are others like me working on hard problems for long periods with the model setting pegged to high. Why wouldn't the companies throttle us? It could even just be that they just apply simple rate limits and that this degrades the effectiveness of the feedback loop between the per…

I run the same config but it tends to fly through those commands on the weekends, very noticeable difference. I wouldn’t be surprised that the subscription users have a (much) lower priority.

That said I don’t go beyond 70% of my weekly limit so there’s that.

Re: Claude Opus 4.5

#436

Post by RestartKernel »

llamasushi wrote: post #30

The burying of the lede here is insane. $5/$25 per MTok is a 3x price drop from Opus 4. At that price point, Opus stops being "the model you use for important things" and becomes actually viable for production workloads. Also notable: they're claiming SOTA prompt injection resistance. The industry has largely given up on solving this problem through training alone, so if the numbers in the system card hold up under a…

> [...] that's legitimately significant for anyone deploying agents with tool access.

I disagree, even if only because your model shouldn't have more access than any other front-end.

Re: Claude Opus 4.5

#437

Post by skylurk »

Aeolun wrote: post #368
Wowfunhappy wrote: post #318

Earlier quoted context omitted.

But again, if that is what the guideline is referring to, why does it say "If you want to emphasize a _word or phrase_". By my reading, it is quite explicitly including single words!

I’m saying that being pedantic on HN is a worse sin than capitalizing a single word. Being technically correct isn’t really relevant to how annoying people think you are being.

I come here for the rampant pedantry. It's the legalism no one wants.

Re: Claude Opus 4.5

#438

Post by threeducks »

idonotknowwhy wrote: post #296
threeducks wrote: post #217

Earlier quoted context omitted.

Rough ballpark estimate: - Amazon Bedrock serves Claude Opus 4.5 at 57.37 tokens per second: https://openrouter.ai/anthropic/claude-opus-4.5 - Amazon Bedrock serves gpt-oss-120b at 1748 tokens per second: https://openrouter.ai/openai/gpt-oss-120b - gpt-oss-120b has 5.1B active parameters at approximately 4 bits per parameter: https://huggingface.co/openai/gpt-oss-120b To generate one token, all active parameters must…

I love logical posts like this. There are other factors like mxfp4 in gpt-oss, mla in deepseek, etc. >Amazon Bedrock serves Claude Opus 4.5 at 57.37 I checked the other Opus-4 models on bedrock: Opus 4 - 18.56tps Opus 4.1 - 19.34tps So they changed the active parameter count with Opus 4.5

Good observation!

56.37 tps / 19.34 tps ≈ 2.9

This explains why Opus 4.1 is 3 times the price of Opus 4.5.

Re: Claude Opus 4.5

#439

Post by zbyforgotp »

jaggs wrote: post #429
lukev wrote: post #154

Earlier quoted context omitted.

There are two possible explanations for this behavior: the model nerf is real, or there's a perceptual/psychological shift. However, benchmarks exist. And I haven't seen any empirical evidence that the performance of a given model version grows worse over time on benchmarks (in general.) Therefore, some combination of two things are true: 1. The nerf is psychologial, not actual. 2. The nerf is real but in a way that…

https://www.youtube.com/watch?v=DtePicx_kFY "There's something still not quite right with the current technology. I think the phrase that's becoming popular is 'jagged intelligence'. The fact that you can ask an LLM something and they can solve literally a PhD level problem, and then in the next sentence they can say something so clearly, obviously wrong that it's jarring. And I think this is probably a reflection of…

There is something not right with expecting that artificial intelligence will have the same characteristics as human intelligence. (I am answering to the quote)

Re: Claude Opus 4.5

#440

Post by Thev00d00 »

esperent wrote: post #346
nickandbro wrote: post #324

"Create me a SVG of a PS4 controller" Gemini 3.0 Pro: https://www.svgviewer.dev/s/CxLSTx2X Opus 4.5: https://www.svgviewer.dev/s/dOSPSHC5 I think Opus 4.5 did a bit better overall, but I do think eventually frontier models will eventually converge to a point where the quality will be so good it will be hard to tell the winner.

I can only see the svg code there on mobile. I don't see any way to view the output.

Click the export tab
Post reply on HN