Deprecating 2 of the only multimodal models with 3 days heads up? Rush in to introduce Gemma 4 model that doesn’t even exist in documentation and register sub 200 tps in my local benchmark? I might as well go back to normal openai api. I thought sambanova is trying to crack fast inference like groq and cerebra’s but the only multimodal is barely usable.
coupled to the 6 months bugs of free tier users upgrading to developer plan and still getting 429 request error and the reverting and re-reverting of usage and payment UI.
I am starting to think this company is ran by vibe coding bros.
/rant