To be honest not as interested as I was to try Opus 4.7 which is now absolute dog muck but Opus 4.8 has dropped if anyone cares to let us know what they think? Supposedly smashed all benchmarks and now is surpassing GPT 5.5...
1 Like
Iām bracing for dog muck platinum, same smell, shinier packaging.
2 Likes
I've found if I start an initial session with Sonnet 4.6 and then after the first couple of responses switch to Composer 2.5 (not set to 'Fast' as that is 5 x the cost of base Composer 2.5), once there is some context, the two models combined absolutely smash it! Such a shame Composer is only available in Cursor. Seriously underrated model. Copilot is absolutely finished from June 1st (Cursor is my current IDE for vibing, with Claude Code and Open Code for CLI purposes). Thus when it comes to even considering Opus 4.8 this combination has to fail first. Then I've always got Minimax and DeepSeek to fall back on which are both equally brilliant for the cost.