ArrmesNeuralRouterV3 (v1.6.0 / v9 Weights)

ArrmesNeuralRouterV3 is the production CPU-native multi-head hashed MLP + structural verifier router for Arrmes (v1.6.0), covering the 20-tool arrmes.tools.v2 contract and the 23-tool arrmes.tools.v3 contract (68 native service_query pairs across 11 services: Sonarr, Radarr, Lidarr, Readarr/Chaptarr, Whisparr, Prowlarr, Bazarr, Tautulli, Seerr, qBittorrent, and Bitmagnet).

Artifact Verification

File: router_weights.npz (7,866,855 bytes / 7.50 MiB)
SHA-256: 17e12b43d64240916b6220e81ba32d863b1f4b6b534c16b2c0c9c559139b7287
Locked acceptance-v7 (Template-Disjoint, N=2,580):
- Positive v3 (1,380 cases, 276 clusters): 1,380/1,380 (100.0%, Wilson 95% LB 99.72%, Cluster-Bootstrap 95% LB 100.0%)
- Positive v2 (1,060 cases, 212 clusters): 1,060/1,060 (100.0%, Wilson 95% LB 99.64%, Cluster-Bootstrap 95% LB 100.0%)
- Hard Negatives (1,200 cases, 240 clusters): 1,200/1,200 (100.0%, Wilson 95% LB 99.68%, 0 unauthorized writes)
Apple M4 Pro CPU Warm Latency: p50 = 0.678 ms, p95 = 0.998 ms (75.38 MiB peak process RSS; zero GPU/Metal allocations).

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Collection including NorviTech/arrmes-neural-router-v3

Arrmes — Local Tool-Calling Router & Verification Suite

Collection

Local tool calling for Sonarr, Radarr and self-hosted media. Arrmes CPU intent router and Qwen3 and FunctionGemma LoRA adapters. • 3 items • Updated 2 days ago