whichllm 0.6.7

A fix for the pack whichllm names under a model. On some PCs it named a pack built for a PC one size bigger, with the speed measured on that bigger PC.

What happened

On my PC, an RTX 5080 with 16 GB and 64 GB of RAM, 0.6.6 named Abyss's 24G64G pack under Qwen3.8-Flash-Next, with the 69.9 tok/s its docs measured on a 24 GB GPU. whichllm engines counted 8 of Abyss's 15 packs where 6 fit. It only showed up once the disk had room for the bigger pack.

Two things added up.

• Abyss's card gave each pack's memory as the GPU plus the memory limit the author tests with, 58 GB for a 64 GB PC. whichllm compares that with all of a PC's RAM, so the PC started 5 GB ahead.
• The 4% allowance from 0.6.5, for a PC that reports under the size it is sold as, was taken off what a pack needs. That also shrank needs that are measured figures and no PC size.

The PC reports 15.9 + 63.3 = 79.2 GB, and the pack's 82 GB less 4% is 78.7.

What changed

• The allowance is now on the machine's side. RAM or VRAM within 4% under a size PCs are sold with counts as that size, so 63.3 GB is 64 and 15.9 is 16. A pack's need is compared as written.
• Abyss's packs give the GPU and RAM of the PC they are named for, together. 24G64G is 88 GB.
• Strata's Coder IQ1_M pack needs 33 GB of RAM and no longer counts as in memory on a 32 GB PC that reports 31.9 GB. It reads from the SSD there, as its docs say.
• Strata's DeepSeek V4.1 packs at 1.75 and 2.25 bpw need 121 and 155 GB. They no longer count as in memory on a 24 GB GPU with 96 and 128 GB of RAM.

Every fit the new rule allows, the old one allowed too, so nothing new starts to fit. If 0.6.6 named a pack for you and 0.6.7 names a smaller one, the smaller one is the one built for your PC.

How it was checked

I compared the new rule with 0.6.6 over 154 PCs, 11 GPU sizes by 14 RAM sizes as they report themselves, for every engine but Abyss. The three Strata packs above are the only ones that change. A new test walks every Abyss pack named for a PC. The pack fits that PC as it reports itself, and doesn't fit in memory on the next PC down. With my PC's sizes whichllm names the 16G64G pack with 70.7 tok/s and counts 6 of 15. All 1,989 tests pass.

Upgrade

:::bash
uv tool upgrade whichllm
pip install -U whichllm
brew upgrade whichllm

Everything else is in CHANGELOG.md (github.com).