latest kimi or glm or qwen coder or minimax? i understand that k2.5 is probably the most challenging one due to its sheer size
2 Likes
A model with diarization can be one the greatest model
I would like to see Qwen 2.5 72B
1 Like
I have sadly had to stop using Groq due to the lack of progress on the available models. Inference speed can’t make up for the lack of a models basic abilities. And as of today almost every model Groq offers is 2 to 3 models behind what is actually usable and worth paying for. Qwen 3.5,GLM 5, Kimi k2.5 any other models that rank near or above opus 4.5
2 Likes
I just want you to upgrade the whisper model; it’s quite bad at misidentifying time and voice, which I find confusing, while other providers do a pretty good job.
Gemma-4 Please!!! 
5 Likes
@yawnxyz gemma-4 is something we are seariously looking forward into!
3 Likes
Gemma 4 + kimi 2.5 plsssssssss
1 Like
Would love to see Googles new Gemma 4 models!
Gemma 4, Kimi2.5, Minimax 2.7(if they are open sourced)
Would love to see Gemma 4 31B on Groq! Its extended context window is a game-changer for managing the load of complex, multi-tool pipelines and deep skill chaining.
1 Like
Please, regarding any Mercury model from Inception Labs — I performed an approximate calculation and estimated it could achieve about 3,200 tokens per second. I just want to clarify the meaning of the term “ridiculously fast.”
Gamma 4. Its about time. I dont want t use Together AI or Cerebras. Groq is the best and fastest.
don’t feel like they’re actually interested anymore. Kimi K2 removed, Minimax 2.5 locked behind enterprise plan, Qwen3 32B somehow also locked behind enterprise plan.
I understand the challenge that is the behemoth Kimi K2/K2.5 on Groq hardware, but Minimax M2.5, and Qwen 3.5/3.6 models (aside from 397MB maybe) should very much be doable?