Thinking Machines’ open 26.4B Inkling 2 Mini loops 12 of its 24 layers three times per token and fits in about 20GB. Most scores are self-reported.
Tag: efficiency
A Hugging Face post runs a 180B open MoE model on a 32GB MacBook by streaming experts from the SSD. It’s about 6% behind a rival engine.
Google’s mid-tier Gemini 4 Neon ranks third on an independent index. Its safety results are mostly inherited, and its own report isn’t out yet.
SenseTime’s open SenseNova-Loop-9B runs most of its layers twice, buying depth with time instead of memory. Its scores are self-reported.
Tencent’s open 220B Hy4 flash keeps 63.5B parameters in phrase lookup tables. It’s very cheap to run, but trails rivals in independent tests.
Tether’s open 0.8B BotPsy calls tools on-device and doubles its base model’s agent scores. Its makers say: keep it from money. Scores are self-reported.
Sarvam’s open Prism writes no text: it scores your options in one pass, in 44 ms, and says how sure it is. The scores are self-reported.
Ricursive says its open, dense Ripple-65B edges out far larger open rivals at terminal coding. The lead is 0.4 points, and self-reported.
IBM’s open Granite Switch 4.1.1 swaps two plug-in parts and leaves the 30B base model untouched. A modular idea, still light on numbers.
Liquid AI’s open-weight LFM3: six ~2B models for text, vision, audio and decisions, grown from one base. The scores are self-reported.