NEC squeezes a 99.9 Wh battery into a 999 g laptop, claims two-day runtime without a charger

[Redaktion]

NEC's 999 g LAVIE BM94C pairs a 99.9 Wh battery with a carbon-reinforced chassis, promising up to 50.8 hours of idle use or 30.5 hours of video playback. The 14-inch business laptop launches in late July with Intel Core Ultra Series 3 processors.

https://www.notebookcheck.net/NEC-squeezes-a-99-9-Wh-battery-into-a-999-g-laptop-claims-two-day-runtime-without-a-charger.1350765.0.html

[But note this about AI]

See, this is a device that stands out. But:

up to 32 GB of LPDDR5X memory

means, if you care about using this laptop for AI extensively, know:

About running AI LLM models locally and privately:
If you plan to get this and run AI, know that up to 32 GB leaves you hardly with any memory for the following SOTA LLM models:
32 GB RAM - 8 GB for the OS itself = 24 GB free for fitting an AI LLM model.

The current SOTA AI LLM models, for their size, are:
huggingface.co/Qwen/Qwen3.6-27B (dense architecture) (Downloads last month: 5.0 million)
and
huggingface.co/Qwen/Qwen3.6-35B-A3B (MoE architecture) (Downloads last month: 5.6 million).

And popular, bang for the buck, quants are:
huggingface.co/unsloth/Qwen3.6-27B-MTP-GGUF UD-Q4_K_XL (17.9 GB)
and
huggingface.co/unsloth/Qwen3.6-35B-A3B-MTP-GGUF UD-Q4_K_XL (22.9 GB) (good for 80k-120k of context).

Additional 2 GB for 32k and 8 GB for 128k (linear)[1] of context are required. This means that the quant barely fits and there's pretty much almost or no space left for any context, see e.g. this:

.. I've come to the conclusion that (1) 32768 is the biggest context I can get away with in an adequately smart model, and (2) it just ain't enough.

In a dense 27B model, all 27B parameters per token are used/activated vs the 35B-A3B, where only 3B are active. Hence, for the same size, a dense model performs much better[2], but runs slower than the 35B-A3B model.

You may want to try the mentioned 27B quant, if you want to have more context.

What would be needed are additional 8 GB of RAM and/or VRAM, either through an 8 GB VRAM dGPU (then it would be basically a gaming laptop), and/or through a total of 48 GB system RAM (RAM amounts go typically up in these steps: 32 GB -> 48 GB -> 64 GB).

[1] reddit.com/r/LocalLLaMA/comments/1tvluaj/how_much_vram_needed_for_qwen_36_27b_q8_with_262k
[2] See 27B dense vs 122B-A10B (MoE) evaluation: huggingface.co/Qwen/Qwen3.5-122B-A10B#benchmark-results. The 122B MoE model requires 4.5 times more memory vs 27B dense, but they both score about the same.

Your own ChatGPT, offline: AI without the cloud on your laptop

notebookcheck.net/Your-own-ChatGPT-offline-AI-without-the-cloud-on-your-laptop.1341872.0.html
-> notebookchat.com/index.php?topic=315954.0.