Claude surpasses the GPT-4 barrier

Recently, Claude 3 was released with three models (Opus, Sonnet, Haiku) which, according to their benchmarks, claim to be better than GPT-4.

It's great to see a company that is capable of competing with GPT-4, as they have practically had the hegemony throughout this entire year.

The prices of the Claude models are very competitive, and they vary quite a bit depending on the model, with Opus being the most expensive and Haiku being much cheaper.

I've been testing them these past few days and I'm very satisfied with their performance. The only drawback is that the Chat part is not yet available in Spain, but its API is.

Prices: (Million tokens, Input / Output)

• Claude Opus (200k): $15 / $75
• Claude Sonnet (200k): $3 / $15
• Claude Haiku (200k): $0.25 / $1.25

The good thing about this variety of models is that you can experiment to see which one is better (or sufficient) for the task you need, and with this, optimize costs.

We can compare them with OpenAI's prices as of today:

• GPT-4 Turbo (128K): $10 / $30
• GPT-4 (8K): $30 / $60
• GPT-4 (32K): $60 / $120
• GPT-3.5 Turbo: $0.50 / $1.50

In my case, I've been using Claude more than ChatGPT for a week, I'll tell you about it in another post. In general, it seems cheaper, works fine and has vision capabilities.