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

Claude Opus 4.6

anthropic.com

Post reply on HN
Print view View original

731–740 of 1001 posts

Re: Claude Opus 4.6

#731

Post by barrell »

surajkumar5050 wrote: post #396

I think two things are getting conflated in this discussion. First: marginal inference cost vs total business profitability. It’s very plausible (and increasingly likely) that OpenAI/Anthropic are profitable on a per-token marginal basis, especially given how cheap equivalent open-weight inference has become. Third-party providers are effectively price-discovering the floor for inference. Second: model lifecycle econ…

> It’s very plausible (and increasingly likely) that OpenAI/Anthropic are profitable on a per-token marginal basis

Can you provide some numbers/sources please? Any reporting I’ve seen shows that frontier labs are spending ~2x on inference than they are making.

Also making the same query on a smaller provider (aka mistral) will cost the same amount as on a larger provider (aka gpt-5-mini) despite the query taking 10-100x longer on OpenAI.

I can only imagine that is OpenAI subsidizing the spend. GPUs cost by the second for inference. Either that or OpenAI hasn’t figured out how to scale but I find that much less likely

Re: Claude Opus 4.6

#732

Post by matt_lo »

ck_one wrote: post #482

Just tested the new Opus 4.6 (1M context) on a fun needle-in-a-haystack challenge: finding every spell in all Harry Potter books. All 7 books come to ~1.75M tokens, so they don't quite fit yet. (At this rate of progress, mid-April should do it ) For now you can fit the first 4 books (~733K tokens). Results: Opus 4.6 found 49 out of 50 officially documented spells across those 4 books. The only miss was "Slugulus Eruc…

use AI to rewrite all the spells from all the books, then try to see if AI can detect the rewritten ones. This will ensure it's not pulling from it's trained data set.

Re: Claude Opus 4.6

#733

Post by fnordpiglet »

legitster wrote: post #72

I'm still not sure I understand Anthropic's general strategy right now. They are doing these broad marketing programs trying to take on ChatGPT for "normies". And yet their bread and butter is still clearly coding. Meanwhile, Claude's general use cases are... fine. For generic research topics, I find that ChatGPT and Gemini run circles around it: in the depth of research, the type of tasks it can handle, and the qual…

Enterprise, government, and regulated institutions. It’s also defacto standard for programming assistants at most places. They have a better story around compliance, alignment, task based inference, agentic workflows, etc. Their retail story is meh, but I think their view is to be the aws of LLMs while OpenAI can be the retail and Gemini the whatever Google does with products.

Re: Claude Opus 4.6

#734

Post by Squarex »

deaux wrote: post #689
Squarex wrote: post #431

Earlier quoted context omitted.

Claude sucks at non English languages. Gemini and ChatGPT are much better. Grok is the worst. I am a native Czech speaker and Claude makes up words and Grok sometimes respond in Russian. So while I love it for coding, it’s unusable for general purpose for me.

You mean Claude sucks at Czech. You're extrapolating here. I can name languages that Claude is better at than GPT. Gemini is the most fluent in the highest number of human languages and has been for years (!) at this point - namely since Gemini 1.5 Pro, which was released Feb 2024. Two years ago.

Yeah, sure, I was overly generalising it from one experience.

Re: Claude Opus 4.6

#735

Post by altmanaltman »

ck_one wrote: post #562
golfer wrote: post #560

Earlier quoted context omitted.

There's lots of websites that list the spells. It's well documented. Could Claude simply be regurgitating knowledge from the web? Example: https://harrypotter.fandom.com/wiki/List_of_spells

It didn't use web search. But for sure it has some internal knowledge already. It's not a perfect needle in the hay stack problem but gemini flash was much worse when I tested it last time.

> But for sure it has some internal knowledge already.

Pretty sure the books had to be included in its training material in full text. It's one of the most popular book series ever created, of course they would train on it. So "some" is an understatement in this case.

Re: Claude Opus 4.6

#736

Post by Squarex »

kuboble wrote: post #509
Squarex wrote: post #431

Earlier quoted context omitted.

Claude sucks at non English languages. Gemini and ChatGPT are much better. Grok is the worst. I am a native Czech speaker and Claude makes up words and Grok sometimes respond in Russian. So while I love it for coding, it’s unusable for general purpose for me.

Claude code (opus) is very good in Polish. I sometimes vibe code in polish and it's as good as with English for me. It speaks a natural, native level Polish. I used opus to translate thousands of strings in my app into polish, Korean, and two Chinese dialects. Polish one is great, and the other are also good according to my customers.

> I sometimes vibe code in polish

This is interesting to me. I always switch to English automatically when using Claude Code as I have learned software engineering on an English speaking Internet. Plus the muscle memory of having to query google in English.

Re: Claude Opus 4.6

#737

Post by zwaps »

hhh wrote: post #397
Someone1234 wrote: post #392

Earlier quoted context omitted.

Specifically including routing (i.e. which model you route to based on load/ToD)? PS - I appreciate you coming here and commenting!

There is no routing with API, or when you choose a specific model in chatGPT.

In the past it seemed there was routing based on context-length. So the model was always the same, but optimized for different lengths. Is this still the case?

Re: Claude Opus 4.6

#738

Post by robertclaus »

tedsanders wrote: post #369
wasmainiac wrote: post #265

Earlier quoted context omitted.

Dumb question. Can these benchmarks be trusted when the model performance tends to vary depending on the hours and load on OpenAI’s servers? How do I know I’m not getting a severe penalty for chatting at the wrong time. Or even, are the models best after launch then slowly eroded away at to more economical settings after the hype wears off?

We don't vary our model quality with time of day or load (beyond negligible non-determinism). It's the same weights all day long with no quantization or other gimmicks. They can get slower under heavy load, though. (I'm from OpenAI.)

Hi Ted! Small world to see you here!

Re: Claude Opus 4.6

#739

Post by wasmainiac »

nl wrote: post #578
wasmainiac wrote: post #566

Earlier quoted context omitted.

Thanks for the response, I appreciate it. I do notice variation in quality throughout the day. I use it primarily for searching documentation since it’s faster than google in most case, often it is on point, but also it seems off at times, inaccurate or shallow maybe. In some cases I just end the session.

Usually I find this kind of variation is due to context management. Accuracy can decreases at large context sizes. OpenAI's compaction handles this better than anyone else, but it's still an issue. If you are seeing this kind of thing start a new chat and re-run the same query. You'll usually see an improvement.

I don't think so. I am aware that large contexts impacts performance. In long chats an old topic will someone be brought up in new responses, and the direction of the mode is not as focused.

Regardless I tend to use new chats often.

Re: Claude Opus 4.6

#740

Post by andkenneth »

bobbylarrybobby wrote: post #729
legitster wrote: post #72

I'm still not sure I understand Anthropic's general strategy right now. They are doing these broad marketing programs trying to take on ChatGPT for "normies". And yet their bread and butter is still clearly coding. Meanwhile, Claude's general use cases are... fine. For generic research topics, I find that ChatGPT and Gemini run circles around it: in the depth of research, the type of tasks it can handle, and the qual…

I really like that Claude feels transactional. It answers my question quickly and concisely and then shuts up. I don't need the LLM I use to act like my best friend.

Weirdly I feel like partially because of this it feels more "human" and more like a real person I'm talking to. GPT models feel fake and forced, and will yap in a way that is like they're trying to get to be my friend, but offputting in a way that makes it not work. Meanwhile claude has always had better "emotional intelligence".

Claude also seems a lot better at picking up what's going on. If you're focused on tasks, then yeah, it's going to know you want quick answers rather than detailed essays. Could be part of it.

Post reply on HN