Run 35B MoE models in ~2.5 GB of RAM. Weights stream from SSD; a trained router predicts loads a step ahead. Mac, iPhone, Android, Windows supported
Python
docs: edge0-10b is 8B-class (Ling 3.0 tiny base)
linyubupad04962b9f532
diff --git a/docs/models/edge0-10b.md b/docs/models/edge0-10b.mdindex 4a66a0c..d86719d 100644--- a/docs/models/edge0-10b.md+++ b/docs/models/edge0-10b.md@@ -1,6 +1,6 @@# edge0-10b-The lightweight, high-throughput tier of the edge0 platform: a 10B-class sparse mixture-of-experts model built on the Ling 3.0 hybrid architecture (MLA + MoE). It trades some parameter count versus the 35b tier for lower peak memory and a higher generation rate, making it a good fit for latency- and memory-sensitive scenarios.+The lightweight, high-throughput tier of the edge0 platform: an 8B-class sparse mixture-of-experts model built on the Ling 3.0 hybrid architecture (MLA + MoE). It trades some parameter count versus the 35b tier for lower peak memory and a higher generation rate, making it a good fit for latency- and memory-sensitive scenarios.The performance profile is based on benchmarks of the current release adapter version and is pinned by default in `Ling10BConfig` (`src/edge0/models/edge0_10b/__init__.py`).@@ -8,7 +8,7 @@ The performance profile is based on benchmarks of the current release adapter ve| Item | Value || --- | --- |-| Parameter scale | 10B-class |+| Parameter scale | 8B-class (≈7.9B total / ≈1.2B active) || Number of layers | 24 (layer 0 is dense) || Number of experts | 128 routed experts + 1 always-resident shared expert || top_k (K) | 8 (native routing width) |