Claude Code and Claude 3.7 Sonnet - Our most intelligent model to date
by•
It's a hybrid reasoning model, producing near-instant responses or extended, step-by-step thinking. One model, two ways to think.
We’re also releasing an agentic coding tool: Claude Code.
For once I was disappointed with Claud. I am impressed by the efficiency and speed of the new model. But the code quality seems to have taken a slight dip. This is only a small test I did to compare with o3-high, grok, Deepseek and claud7 (feel like this name suits it better). Same prompt. `Create Tetris in HTML, css and js` . all 3 other models had a functional game with around 250 lines of code. Claude on the other hand had 600+ lines of code and unfortunately did not run correctly.
Report
For the first time I'm underwhelmed with a Claude release. I'm a huge fan since your first beta and refuse to use other models, but what always made you my number 1 in the past is missing in the new release: the biggest context window on the market! Gemini has done the right step - increase the input tokens! By now this is the only real differentiator and what makes a LLM truly individual.
Apart from the coding capabilities, I can sense the reduction in reasoning errors. Well done!
Report
Discriminations a lot by this company👎🏻
Report
Obsessed.
Report
Has to be one of the best out there .
Report
Absolutely love using Claude 👏 Great product all round, especially use it for writing!
Report
The jump from 3.5 to 3.7 is honestly bigger then i expected. the extended thinking mode is where it really shines. I gave it a gnarly refactoring task that would've taken me half a day and it just worked through it step by step and got it right. claude code + 3.7 together is a genuinely different experience from anything else out there rn.
Replies
Minara
Amazing tool! Best tool I've ever used, save so much time for me. Love it! If make the context size larger would be much much better!! Thanks team!
Minara
realy grateful for this tool
Quadratic
We're implementing Claude 3.7 into our product (Quadratic.ai) today! It's performing so much better than 3.5
@cole_at_quadratic This looks promising, I love the simple UI :::
For once I was disappointed with Claud. I am impressed by the efficiency and speed of the new model. But the code quality seems to have taken a slight dip. This is only a small test I did to compare with o3-high, grok, Deepseek and claud7 (feel like this name suits it better). Same prompt. `Create Tetris in HTML, css and js` . all 3 other models had a functional game with around 250 lines of code. Claude on the other hand had 600+ lines of code and unfortunately did not run correctly.
For the first time I'm underwhelmed with a Claude release. I'm a huge fan since your first beta and refuse to use other models, but what always made you my number 1 in the past is missing in the new release: the biggest context window on the market! Gemini has done the right step - increase the input tokens! By now this is the only real differentiator and what makes a LLM truly individual.
Scene
Apart from the coding capabilities, I can sense the reduction in reasoning errors. Well done!
Obsessed.
Has to be one of the best out there .
Absolutely love using Claude 👏 Great product all round, especially use it for writing!
The jump from 3.5 to 3.7 is honestly bigger then i expected. the extended thinking mode is where it really shines. I gave it a gnarly refactoring task that would've taken me half a day and it just worked through it step by step and got it right. claude code + 3.7 together is a genuinely different experience from anything else out there rn.