I strongly recommend the Legion Pro series. Check gaminglaptop.deals against brand quality. Current favorite deal: Legion 5i 4070 $1399 (as of 2025-03-25, may change before I update). Also get this cooling pad to control temps (which impacts performance) and protect from dust.
If you're buying it only for gaming, get as high a GPU as you can afford in the Nvidia 40-series, anything 4070 and above. The 50-series raster gains are marginal (low bang-for-buck). And DLSS4 (the main performance marketing of the 50-series launch) is backwards-compatible to 40. This offers AI upscaling and multi-frame generation, which improve gaming performance significantly. It opens performance improvements to software instead of hardware, so your laptop can continue to improve as Nvidia drivers release. You'll be limited in DLSS capabilities if you buy a 30-series Nvidia; so if you plan to game, get a 4070+. I recommend a 4080. Nvidia is recommended over AMD due to DLSS.
If you're buying the laptop for ML training and gaming, get 4090. With ML, VRAM is more essential than compute, and the top tier of each series (xx90) has the highest VRAM. Nvidia is required - ML frameworks are Nvidia-first, through CUDA and CuDNN.
If you're buying the laptop only for work in ML, you can downgrade to 3090 and save lots of money. Much cheaper than 40/50-series, but in the VRAM sweet-spot at an xx90. Use WSL2 (Ubuntu) + Docker for your ML work, see mla/12 for details.
Why not Mac? For gamers it's obvious. But for ML, while strides are being taken in Apple Silicon (eg with LLMs), it's still not first-class. If you plan to do heavy ML work, you're much safer with a PC.