We’re Pipe Network, a decentralized CDN for Web3 content .
Join our global mesh of hyper‑local PoP nodes and earn rewards for helping power the network. Learn more at https://docs.pipe.network
PipeNetwork
Pinned Loading
Repositories
Showing 10 of 25 repositories
- glm53-mlx Public
MLX (Apple Silicon) runtime + quantization for GLM-5.3 (744B glm_moe_dsa): fixed shared-indexer layers, validated against transformers
- hi Public
- glm53-flash-mlx Public
MLX (Apple Silicon) runtime + quantization for GLM-5.3-Flash, validated against transformers
- qwen38-flash-next-mlx Public
MLX (Apple Silicon) runtime + quantization for Qwen3.8-Flash-Next, validated against transformers
- qwen38-mlx Public
MLX port of Qwen3.8-2.4T-A95B (2.4T params, 512-expert MoE) with streaming quantization, REAP expert pruning, and per-layer evaluation for a model too large to run
- muse-glimmer-mlx Public
MLX (Apple Silicon) port of Muse-Glimmer-30B. Validated against transformers; also a runtime for the existing MLX conversions, which no released mlx-vlm can load.
- minimax-h3-mlx Public
MLX (Apple Silicon) port of MiniMax-H3 — 33B joint video+audio diffusion. Validated against the diffusers reference; AdaLN precompute drops 13B at inference.
- kimi-k3-mlx Public
MLX port of moonshotai/Kimi-K3 (2.78T multimodal MoE): streaming converter, REAP expert pruning, and per-language expert-overlap analysis
- deepseek-v4-mlx Public
MLX (Apple Silicon) port of deepseek-ai/DeepSeek-V4-Flash-0731 (304B MoE). From-scratch implementation of Hyper-Connections, hash-routed experts, learned KV compression and sparse-attention indexing — none of which exist in any released runtime.
- pipe-wrap Public
People
This organization has no public members. You must be a member to see who’s a part of this organization.
Top languages
Loading…
Most used topics
Loading…