I can tell in my limited testing the new "optimized" version of grok-code-fast-1 is better than the previous version. It produced results more consistent with other recent models, which makes me think the model was updated with new data/training.
I have never seen a model worse than Grok Code Fast 1 Optimized. It's looping its own thinking and always calls tools incorrectly. Not pleasant to use it.
Cool, nice to see this model return, and interesting because if you look at the Kilo leaderboard over the last couple of months you will notice the the overall token usage went down since Grok code Fast 1 went away, and also most usage by far are the free models, no matter who is supplying them! But thanks guys, nice to see more affordable and free models available again.
Please remember that this comment section is governed by our Kilo Code Code of Conduct https://github.com/Kilo-Org/kilocode/blob/main/CODE_OF_CONDUCT.md
Wow! What a great surprise for a Monday. Time to switch back from Claude.
Great model just like the og - feels like a boost on more complex tasks
I can tell in my limited testing the new "optimized" version of grok-code-fast-1 is better than the previous version. It produced results more consistent with other recent models, which makes me think the model was updated with new data/training.
I have never seen a model worse than Grok Code Fast 1 Optimized. It's looping its own thinking and always calls tools incorrectly. Not pleasant to use it.
Cool, nice to see this model return, and interesting because if you look at the Kilo leaderboard over the last couple of months you will notice the the overall token usage went down since Grok code Fast 1 went away, and also most usage by far are the free models, no matter who is supplying them! But thanks guys, nice to see more affordable and free models available again.