Q: By the way, which BYOK models are suitable for identifying a specific number?
The GPT-4o mini LLM performs well, but I was wondering if there is an option to filter or recognize digit numbers specifically. I tested Pismo, which works well for language processing or recognizing letters a-z, but it struggles with understanding numbers like ascending or descending "12345.", "111." When I asked Pismo to identify a specific number, it failed regardless of how I phrased the prompt.
Also, if I downgrade to the Tier 1 plan, could you please explain how credit consumption is calculated?
By the way, which BYOK models are suitable for identifying a specific number?
Pavel_Pismo
Edited Aug 29, 2025A: Hi, thanks for your question!
Identifying specific numbers is a known limitation of current LLMs. Theyâre optimized for natural language, so sequences like 12345. or 111. can be inconsistent. This isnât fully fixable and isnât model-dependent.
We regularly test different models, including GPT-5, but 4.1-mini currently gives the best results for text and translations - we switched back from GPT-4.1 nano because of reliablility.
And as for Tier 1 credit consumption - it's pretty straightforward. 1 executed prompt equals 1 credit spent.