Claude Opus 4.5 vs 4.6 optimized for Jetbrains June
I wonder if 4.6 is actually better or not? Previous Junie version had 2 Claude Opus available: Claude Opus 4.5 (Fast) - optimized and Claude Opus 4.5 (recommended for high complexity tasks). Opus 4.5 was consuming 5-10x more credits than default, while 4.5 optimized: 2-3x more than default.
After recent Junie update, there's no Opus 4.5 anymore. Instead, we get 4.6 optimized version and it consumes 2-3x more credits than defualt. There's also previous 4.5 optimized version available too.
So, which is actually better for very complex tasks: previous (non optimized) 4.5 or optimized 4.6? As 4.6 is optimized, that's actually update for 4.5 optimized, so I gues it will be worse than previous, regular 4.5, which also consumes more credits for complex tasks, is that true? It seems 4.5 - fast has been updated to 4.6 - fast. But 4.6 regular for high complexity tasks is missing, as well as previous 4.5 for high complexity tasks has been removed.
Is it worth to switch to recent Junie then? I can accept consuming more credits ,but I want high complexity taksks to be solved best.
Please sign in to leave a comment.
Hey, thanks for the question.
According to our rigorous evaluations, Opus 4.6 (optimized version) is clearly the best Anthropic's LLM regarding the quality. And yep, it's significantly better than deprecated Opus 4.5 (non-optimized). Opus 4.5 (optimized) is left mostly for specific setups within b2b segment where the new model is not available. Once you're using Opus, we strongly recommend Opus 4.6
Found this post because I'm wondering the same thing. They seem to be selling us a worse version of the model?
Images to explain the above
Hello, can someone explain this change? What is optimized exactly? Why is so much cheaper? Are we getting a worse version or the same with optimized cost?
I already see people swapping to Cursor just because of this nonsense
Iee2020031 as I mentioned above Opus 4.6 optimized provides the best quality across all models (and yep, it's better than previous non-optimized version of Opus 4.5 which was far more expensive). Optimizations touch model's hyper parameter tuning and prompting strategies. Since we don't have non-optimized Opus in newer Junie version, we're going to drop “optimized” to reduce confusion