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

Claude Sonnet 4.5

anthropic.com

Post reply on HN
Print view View original

351–360 of 819 posts

Re: Claude Sonnet 4.5

#351

Post by sixothree »

CuriouslyC wrote: post #9
yewenjie wrote: post #5

Looking at the chart here, it seems like Sonnet 4 was already better than GPT-5-codex in the SWE verified benchmark. However, my subjective personal experience was GPT-5-codex was far better at complex problems than Claude Code.

The Anthropic models have been vibe-coding tuned. They're beasts at simple python/ts programs, but they definitely fall apart with scientific/difficult code and large codebases. I don't expect that to change with the new Sonnet.

You definitely need some context management like Serena.

Re: Claude Sonnet 4.5

#352

Post by jjordan »

drak0n1c wrote: post #320
beefnugs wrote: post #248

Earlier quoted context omitted.

Still has to be a local model too. Arbitrary government censorship on top of arbitrary corporate censorship is a hell no for me forever into the future

For what you're looking for, VeniceAI is focused entirely on privacy and making their models uncensored. Even if it's not local. They IP block censorious jurisdictions like UK, rather than comply.

VeniceAI is great, and my go-to for running open source models. Sadly they appear to have given up providing leading coding models, making it of limited use to me.

Re: Claude Sonnet 4.5

#353

Post by sixothree »

xwowsersx wrote: post #62
mohsen1 wrote: post #14

Price is playing a big role in my AI usage for coding. I am using Grok Code Fast as it's super cheap. Next to it GPT-5 Codex. If you are paying for model use out of pocket Claude prices are super expensive. With better tooling setup those less smart (and often faster) models can give you better results. I am going to give this another shot but it will cost me $50 just to try it on a real project :(

Same here. I've been using GCF1 with opencode and getting good results. I also started using [Serena]( https://github.com/oraios/serena ), which has been really helpful in a large codebase. It gives you better search than plain grep, so you can quickly find what you need instead of dumping huge chunks of code into Claude or Grok and wasting tokens.

Serena really does feel like a secret weapon sometimes.

Re: Claude Sonnet 4.5

#354

Post by cloverich »

dirkc wrote: post #173
cloverich wrote: post #108

Please y'all, when you list supportive or critical complaints based on your actual work, include some specifics of the task and prompt. Like actual prompt, actual bugs, actual feature, etc. I've had great success with both ChatGPT and Claude for years, am around 3x sustained output increase in my professional work, and kicking off and finishing new side projects / features that I used to simply not ever finish. BUT t…

Would you say you do things you'd normally do 3 times faster? Or does it help you move past the things you'd get stuck on or avoid in the past, resulting in an overall 3x speedup?

Things I'd normally do 3x faster. That 3x is me focusing explicitly on the precise things I did before - the PR rate on a specific work project - because I tie those PR's back to specific tasks the same as I did before I used claude code. I haven't looked at lines of code, total commits, etc. Qualitatively I write more tests and abstract more components than I used to, but those get lumped in to the PRs as I normally try to limit pure refactoring work, and instead tie it into ticketed feature requests or bugs.

I don't count the things I'm doing now that I would have avoided or never finished in the past. For those, of course to me personally those are worth much more psychologically than 3x, but who knows if it's an actual boost. I.e. I took a partially scripted task the other day and fully automated it, and also had it output to the CLI in a kind of dorky sci-fi way because it makes it fun to run it. It didn't take long - 30 minutes? But I certainly didn't _gain_ time doing that, just a little more satisfaction. TBH I'm surprised 3x is so controversial, I thought it was a really cool and far more practical assessment than some of these 10x claims I'm seeing.

Re: Claude Sonnet 4.5

#355

Post by netdur »

peterdstallion wrote: post #157

I am a paying subscriber to Gemini, Claude and OpenAI. I don't know if it's me, but over the last few weeks I've got to the conclusion ChatGPT is very strongly leading the race. Every answer it gives me is better - it's more concise and more informative. I look forward to testing this further, but out of the few runs I just did after reading about this - it isn't looking much better

Yeah, ChatGPT is better overall, but with a twist, Gemini can actually be the best if you use AI Studio, tweak the config, and set up a good system prompt, kinda like how nano banana is SOTA, but Qwen-Edit feels more useful since it’s less censored, meanwhile ChatGPT is starting to feel slower and kinda showing its age, another example is Veo 3 being SOTA while the infamous Grok is technically worse but doing better, and OpenAI’s Sora is pretty much dead

edit: I run low profile service that localizing e-commerce photos, like taking Alibaba listings and swapping the model to look local alike, with nano banana I can’t automate it because I have to manually check if the output got blocked (anything with female skin is risky, underwear or cleavage is 100% blocked), but Qwen-Edit just does the job without fuss

Re: Claude Sonnet 4.5

#356

Post by GoatInGrey »

darksaints wrote: post #85
rudedogg wrote: post #70

I just ran this through a simple change I’ve asked Sonnet 4 and Opus 4.1, and it fails too. It’s a simple substitution request where I provide a Lint error that suggests the correct change. All the models fail. I could ask someone with no development experience to do this change and they could. I worry everyone is chasing benchmarks to the detriment of general performance. Or the next token weight for the incorrect c…

> I worry everyone is chasing benchmarks to the detriment of general performance. I've been worried about this for a while. I feel like Claude in particular took a step back in my own subjective performance evaluation in the switch from 3.7 to 4, while the benchmark scores leaped substantially. To be fair, benchmarking has always been the most difficult problem to solve in this space, so it's not surprising that benc…

Not that it was better at programming, but I really miss Sonnet 3.5 for educational discussions. I've sometimes considered that what I actually miss was the improvement 3.5 delivered over other models at that time. Though since my system message for Sonnet since 3.7 has been primarily instructing it to behave like a human and have a personality, I really think we lost something.

Re: Claude Sonnet 4.5

#357

Post by simonw »

knowsuchagency wrote: post #303
Redster wrote: post #268

Earlier quoted context omitted.

Simonw is a cheerful and straightforward AI journalist who likes to show and not just tell. He has done a good job aggregating and documenting the progress of LLM tools and models. As I understand it, OpenAI and Anthropic have both wisely decided to make sure he has up to date info because they know he'll write about it. Thanks for all your work, Simon! You're my favorite journalist in this space and I really appreci…

I like Simon, but he's not a journalist. A journalist would not have gone to OpenAI to glaze the GPT-5 release with Theo. I don't say this to discount Simon -- I appreciate his writing and analysis but a journalist, he isn't.

I don't call myself a journalist, partly because no publication is paying me to do any of this!

If I had an editor I imagine they would have talked me out of going to the OpenAI office for a mysterious product preview session with a film crew.

Re: Claude Sonnet 4.5

#358

Post by fluidcruft »

iagooar wrote: post #258

Anecdotal evidence. I have a fairly large web application with ~200k LoC. Gave the same prompt to Sonnet 4.5 (Claude Code) and GPT-5-Codex (Codex CLI). "implement a fuzzy search for conversations and reports either when selecting "Go to Conversation" or "Go to Report" and typing the title or when the user types in the title in the main input field, and none of the standard elements match, a search starts with a 2s de…

I've tried codex with GPT-5 a little bit and I haven't figured out how to get it to not be annoying. codex just constantly tries to gaslight and argue with me. For example, I was debugging an OpenGL render pipeline that went black and codex insisted it must be because I was ssh'd into a headless server. It really makes me appreciate the Claude "You're absolutely right!"s. Anyway as you can tell, I haven't cracked working with codex. But at the moment it just messes everything up and ways I've learned to work with claude don't seem to translate.

Re: Claude Sonnet 4.5

#359

Post by sixothree »

mohsen1 wrote: post #14

Price is playing a big role in my AI usage for coding. I am using Grok Code Fast as it's super cheap. Next to it GPT-5 Codex. If you are paying for model use out of pocket Claude prices are super expensive. With better tooling setup those less smart (and often faster) models can give you better results. I am going to give this another shot but it will cost me $50 just to try it on a real project :(

I just can't bring myself to get over the grossness factor of using an x branded product.

Re: Claude Sonnet 4.5

#360

Post by throwaway314155 »

Kuinox wrote: post #64
mhh__ wrote: post #39

Earlier quoted context omitted.

Memorise what exactly?

Coordinate and shape of the element used to form a pellican. If you think about how LLMs ingest their data, they have no way to know how to form a pellican in SVG. I bet their ability to form a pellican result purely because someone already did it before.

> If you think about how LLMs ingest their data, they have no way to know how to form a pellican in SVG.

It's called generalization and yes, they do. I bet you could find plenty of examples of it working on something that truly isn't "present in the training data".

It's funny, you're so convinced that it's not possible without direct memorization but forgot to account for emergent behaviors (which are frankly all over the place in LLM's - where you been)?

At any rate, the pelican thing from simonw is clearly just for fun at this point.

Post reply on HN