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

Claude Opus 4.7

anthropic.com

Post reply on HN
Print view View original

981–990 of 1001 posts

Re: Claude Opus 4.7

#981

Post by port11 »

berkes wrote: post #864
paulluuk wrote: post #836

Earlier quoted context omitted.

If open weight models are sufficient for your engineering problems, then you should absolutely use them. But I haven't seen a single open weight model that can get even close to the complexity in my projects. They sometimes work for small toy examples or leetcode puzzles, but not very any real project. Really curious what models you've found that could replace current state of the art.

I've been using devstral2 with great success for a few months now. The hosted version, not running one locally or such. Devstral is open. Devstral is good, Opus better. But not much. For me, "good" is "good enough". The difference, IME lies in context engineering: skills, agents.md, subagents, tools, prompts. A Devstral with good skills performs far better than an "blank" claude code. Claude with good skills performs…

Perhaps. I’d like to like Devstral because I’d rather give my money to an European business.

My experience with it in an existing codebase has been that it gets to results much more reliably than Gemini Flash or Haiku, but it will cut corners and write incomprehensible code even with a good Opus plan to boot.

It’s true that the context and tooling might help, but setting everything up and finding the arcane mix of correct MCPs/skills is a job in itself right now. What I do see is that I’ve wasted months trying to get good code out of Gemini, Devstral2, and a good experience out of stuff like OpenCode and everything under the sun.

Re: Claude Opus 4.7

#982

Post by slava_vechir_2 »

Opus 4.7 seems a little bit better then Opus 4.6, but I honestly think, that for the fact that it consumes a lot more usage, it is not worth it, especially with the tiny limits you get, even if you are a Pro user.

Re: Claude Opus 4.7

#983

Post by dwood_dev »

michaelcampbell wrote: post #957
dwood_dev wrote: post #736

Earlier quoted context omitted.

I use both. I avoided codex in late 2025 because it was slow as molasses. I tried it again in February and it was on par with Opus speed. I like codex(gpt-5.4 high) more for its ability to nitpick my PRs and find bugs. I like opus 4.6 much better for anything dealing with visuals, but I feel its rule adherence is inferior and it is not nearly as thorough on code reviews. I like working and building better with claude…

> I tried it again in February and it was on par with Opus speed. I'm not finding that, like not even close. I'm using it mostly to write specs and docs and having Claude and Gippity check each other's work and fix things. It's looking into other previously written MD docs, and checking against 3-4 multi-thousand file Ruby codebase(s). 5.4 takes about 50% longer, almost without fail. I'm using 'medium' effort on both…

Might be a difference in use cases. Mostly working in repos under 100k LOC, most of those are under 30k. Primarily golang, python, and terraform. Also a lot of interactive troubleshooting of Kubernetes, GHA, hardware debugging and application debugging.

I also only run in fast mode on gpt-5.4 high.

Re: Claude Opus 4.7

#984

Post by merlindru »

trueno wrote: post #455

noticing sharp uptick in "i switched to codex" replies lately. a "codex for everything" post flocking the front page on the day of the opus 4.7 release me and coworker just gave codex a 3 day pilot and it was not even close to the accuracy and ability to complete & problem solve through what we've been using claude for. are we being spammed? great. annoying. i clicked into this to read the differences and initial exp…

one thing to keep in mind is that you have to use GPT-5.4 differently from codex. they "work" in different ways. i was aghast when i noticed how terrible Codex was against Claude Code only to conclude it was me who wasn't using it right a couple days later

Opus 4.6 and 4.7 are better than GPT-5.4 xhigh, but only marginally. I can't give proper pointers on what to change because it's incredibly hard to quantify.

In essence, though, GPT-5.4 needs explicit instructions not to take liberties - this is included in the default system prompt of Claude Code which leads me to think Opus is just as overzealous as GPT-5.4 unless explicitly told off.

And it takes EVERYTHING you say at face value. Questions like "don't you see why this is bad?" will be answered with "yeah, i do." which is also kind of cool...

because with Opus in Claude Code i constantly have to reassure the model i'm not insinuating anything, lest it takes my question and runs with it into a frenzy of "oh shit my bad let me fix it im so sorry" type changes.

Re: Claude Opus 4.7

#985

Post by big-chungus4 »

hijodelsol wrote: post #855
big-chungus4 wrote: post #848

Crazy how popular this post is on HN, are this many people actually using expensive paid models? Is everyone on HN a millionaire? Or is someone botting all anthropic posts?

I mean, the 100$ plan is less than the hourly rate of any consultant / senior dev in developed countries. So if it can save even one hour a month, it's cost efficient for the customer (at the current, subsidized rates, of course).

So are the majority of people on HN senior devs from a developed country, who like using LLMs for coding? I find that hard to believe

Re: Claude Opus 4.7

#986

Post by big-chungus4 »

tossandthrow wrote: post #852
big-chungus4 wrote: post #848

Crazy how popular this post is on HN, are this many people actually using expensive paid models? Is everyone on HN a millionaire? Or is someone botting all anthropic posts?

200USD a month really is not that much. Especially not for an employer who is used to pay 150-250k a year for an engineer. Especially for the value it provides.

I don't think that the majority of HN users are employers who are used to pay 150k-250k a year

Re: Claude Opus 4.7

#987

Post by phreack »

This has been the worst upgrade so far. Claude Code had been doing great for months, then the past week took a nosedive. And today I find that _continuing_ a session from yesterday that had nothing to do with cybersecurity (literally pasted a stacktrace from a rare crash and told it to help me find a reproduction case to be able to fix it, as we very regularly did) suddenly ran afoul of usage policies and stopped the chat entirely. It's kind of a joke phrase by now, but in this case it's 100% serious, such behavior has made Claude Code literally unusable.

As a bonus, it somehow ate my entire daily allotment in a single prompt, something which had never happened before. I'll try again on Monday and if there's no change cancel my subscription outright and demand a refund.

Re: Claude Opus 4.7

#989

Post by Eifert »

XCSme wrote: post #731
simonw wrote: post #203

I'm finding the "adaptive thinking" thing very confusing, especially having written code against the previous thinking budget / thinking effort / etc modes: https://platform.claude.com/docs/en/build-with-claude/adapti... Also notable: 4.7 now defaults to NOT including a human-readable reasoning token summary in the output, you have to add "display": "summarized" to get that: https://platform.claude.com/docs/en/build-…

The reasoning modes are really weird with 4.7 In my tests, asking for "none" reasoning resulted in higher costs than asking for "medium" reasoning... Also, "medium" reasoning only had 1/10 of the reasoning tokens 4.6 used to have.

Medium reasoning has regressed since 4.6. While None and Max have improved since 4.6 in our benchmark. We suspect that this is how Claude tries to cope with the increased user base. Note, Google and OpenAI probably did something similar long ago.

Re: Claude Opus 4.7

#990

Post by hirako2000 »

algoth1 wrote: post #913
hirako2000 wrote: post #880

I can understand the wishes to make LLMs even more self driven. After all that's the idea of a lose prompt. No matter how short, LLM figures out what most users are expecting. Thanks to RLHL it accomplishes wonders. My desire though is to be able to steer the model exactly where I want. Assuming token cost isn't an issue, it doesn't remove the need for costly review. I would rather think first and polish up my abilit…

I suspect this is part of the reason why gemini 3.1 pro is insanely good on AiStudio and pretty bad on the gemini app. I have thousands of small videos to convert to detailed descriptions and I'm using a super detailed system prompt. It works perfect either via api or Aistudio. I tried doing a gem on the gemini app using the same prompt as the gem instructions and I just can't get the same results. So, the issue migh…

I didn't even know they injected a system prompt into chat apps.
Post reply on HN