Claude is still the best coder

Discussion
1

The backend (java) for all the content below, was completed in only 2 conversations Shocked me, more than 20 files （java+ xml）, and all the features were tested without problems, Claude is simply a god. In the rules file, I will write global documentation, interface paths, request encapsulation, naming conventions, date formats, language comment specifications, and other global content

I will use Claude coding o4mini to solve complex problems on a daily basis!

To tell you the truth, Gemini’s large file capability is very good, but he’s very inked A lot of nonsense, like to write useless comments, I do not know what you think of it

Add new financial information (news, market trends, financial knowledge)
Edit existing information
View information details
Archive information (take offline but retain records)
Restore archived information
Delete information (permanent deletion)
Information status management (published, draft, archived)
Information categorization and filtering
View by category (financial news, market trends, financial knowledge)
Filter by status (published, draft, archived)
Filter by date (today, this week, this month)
Information search function

Market monitoring management
Exchange rate information management (view/edit exchange rates for various currencies)
Interest rate information management (view/edit term deposit rates for different durations)
Stock index management (view/edit indices such as SSE, SZSE, HSI, etc.)
Real-time market data updates

Information publishing control
Publish immediately
Schedule publishing
Send notifications upon publishing

Data statistics display
Total information statistics
Published information statistics
Draft statistics
Archived information statistics

4 Likes
2

To be honest, I don’t know how to code, but Claude in Cursor is incredibly useful—I’m amazed! And the Tab feature is great. The competitor, Windsurf, I purchased the pro, experience was like shi t, same model, he had problems often

1 Like
3

How much do you pay?

1 Like
4

I’m using normal claude thinking I’m not using max.

1 Like
5

OP: are you using Claude 3.5 or 3.7: thinking or non-thinking? I have a large PHP project and I concur, I prefer Claude over Gemini.

2 Likes
6

3.7 thinking

1 Like
7

I’ve been using Gemini quite a lot in the last few weeks, I used Claude exclusively before that. I feel Gemini has more common sense. But a worse engineer e.g. useless comments – I agree.

1 Like
8

In the new version of the gemini model, it’s really good for context and speed. It’s just that I think he has way too much cra p. Saying a bunch of stuff and writing a bunch of comments. And it’s just fast, but it actually takes longer to think, up to thinking for a few minutes counting the time spent thinking, it’s not as fast as claude.

2 Likes
9

Is this the only model ticked or other models are ticked as well?

I’ve not touched mine since I have had this

image
image606×639 4.22 KB

I only use it for Android Programming.

1 Like
10

Hi, can you elaborate on your question? Didn’t understand what you meant. From the picture it looks like you have chosen models that you can check off in the dialog. the claude model is still very nice and it has a very standardized comment. It’s a good engineer, just more active.

11

In your model list is Claude the only model ticked for you or do you’ve others ticked?

12

I only checked off claude 3.7 and 3.7 thinking and o4mini

1 Like
13

Hi

Any knows how the economy of using MAX contra not using MAX is?
In the aspects of Time, Money, Quality.

14

Except for the context, and the number of tool calls. Nothing else seems to have changed. The max price is probably more than ten times the normal one. Depends on how much you use it. I’ve used max a few times and each conversation request consumed more than 10 quick requests Used o3max once and it consumed 100 quick requests for me For everyday situations, just use the regular model. Didn’t see much difference in quality.

2 Likes
15

Yes, it becomes very expensive right of the rail.

16

Have you used 3.5 before, and which one is better? Half a year ago, people were certain that 3.5 was still better at that time.

17

To me, 3.7 is better, it’s smarter, it can handle more tasks at once. I used cursor last October and 3.5 had a hard time with 1000 lines of code or 10 files.

If you need a model that follows your commands Just use gpt4.1.

1 Like
18

I too keep coming back to Claude after giving all highly ranked newcomers a try or two. I find it extremely predictable relative to the others, and just as powerful for the languages I am using (Python, C++, bash, CLI, LaTex, MarkDown, …). For best results, I keep my files small (well under 1200 lines or so), because the Cursor editor tool tends to get confused and damaging in long files (Claude is great at moving functions out to separate helper files). I also have Claude maintain a status and to-do document as we go along, because the automatic summary and reboot is not nearly as complete (neither Cursor’s or Claude Code’s). I never use Max because it already thinks plenty for the clearly specified incremental tasks I am giving it. I try to single-step everything it writes at least once, and I ask it about unfamiliar constructs. Finally, I have it write documentation and unit tests, and lately there has been very little for me to polish further. The work has been superb, and I hear we get a new model late this week.

1 Like
19

Yes, claude is still very good , comment specification, maintainability is very good, I hope to be able to launch claude 4 large file editing failures are basically slow, long time editing is not finished cursor will determine the timeout, which can not be helped, but the new version of the cursor optimization of this I use claude easily edited 2000 lines of documents (large file refactoring is not working, it ■■■■■, but eventually it will be fixed)

20

I would say that Claude is not always the best coder, it is definitely the most creative. gemini 2.5 (newest update) is now amazing at debiging and writing code.
claude is still the most stable.