Observed arrival · 2026-10-25
talachu’s small-model field guide
A curated guide to open-weight small AI models, with an interactive calculator for estimating their memory needs on different hardware.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- People choosing small open-weight models to run locally or self-host
- Worth noticing
- The calculator includes KV-cache precision, context length, concurrent users, and hardware memory-bandwidth presets.
Field notes
The calculator distinguishes model memory from real inference memory, including a KV cache that grows with context and concurrent users. Its hardware presets span several CPU memory bands, Apple silicon tiers, and RTX cards; one displayed example estimates about 5 GB total on a 24 GB setup. The model list is organized by modality, and its latest visible item is EmbeddingGemma 2, described as a 740M-parameter Apache 2.0 model.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
✦PrettyNotable craft visible
●ProPolished or operationally mature
◎NicheUnusually specific use
One card from the complete issue