Testing only LLMs
After buying Tier 3 I have refund it because:
- I was unable to get a straightforward French to English translation, as it kept asking how I’d like to proceed and repeatedly offered to summarize the text. Even though I specified that it has occurred with different models Straico's answer was: "I’d strongly recommend testing multiple models to find the one that provides the best results for your specific task." But what if I like how a specific model translates?
- Another issue is that I requested a complete translation in one response - as the model offered it itself, but it did not initially provide this. However, after translating in parts, I asked it to combine the text, and it did, indicating that the issue lies not with the text length but with something else. Straico's answer: "it’s worth noting that AI providers often enforce character or word limits on the model’s outputs." "We recommend you try with multiple models to determine which model has a longer output limit, so you can get the answer in fewer outputs. "
- Additionally, Straico's support also mentioned, "the issue doesn’t appear to be on our end", so I tested the same prompt in the LLM's original app, and surely the response was accurate and concise, free of unnecessary questions, opinions, or divided answers.
So, I believe that Straico uses a pre-prompt that needs adjustment, as it complicates the user's tasks, and consumes time and credits on jobs the user did not request. And even though Straico's team will have a positive attitude in their responses, the real outcome you get from Straico - at least for now - is that the LLMs on their app are for testing only, at the price of your time and credits.
SantiagoC_Straico
Jan 6, 2025Hello Booky!
Thank you for sharing your detailed feedback. We take your experience very seriously and appreciate the opportunity to address the points you’ve raised, as this helps us grow and improve Straico for everyone.
First, regarding translations: we understand that achieving accurate and reliable AI translations can be challenging, and we’d love to learn more specifics about the issues you encountered. Your insights would allow us to refine our system further. To support this, we're actively working on enhancing our LLM selector, which will enable users to better match their needs with the most suitable model for their specific tasks.
We’ve also noticed the "pre-prompt" concern you mentioned. To clarify, when Smart Mode is turned off, the system processes your input exactly as written, without any adjustments or alterations. However, when Smart Mode is on, it optimizes the prompts for specific use cases to help deliver better results. This distinction ensures that users can decide how much guidance or optimization they’d like in their interactions.
It's also worth noting that we exclusively use official models provided by trusted API providers like OpenAI, Anthropic, and Perplexity. These are not test versions, but the same high-quality production models relied upon by enterprise customers, ensuring reliability and performance.
We’re truly sorry to hear that your experience with Straico has been disappointing so far. We acknowledge that we may not always get it right for every user, and that’s where your feedback is so valuable. Please feel free to reach out to our support team for personalized assistance, we’re here to help address any remaining concerns or challenges you may have.
Thank you again for providing your thoughts, as they are crucial to helping us evolve and improve Straico for all users.