YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

G1L Knot Archive (non-production)

This repository is the non-production archive for Forkjoin.ai knot artifacts. Nothing here is a serving artifact. Production knots live in R2 under distributed-inference/models/ and are referenced only through admission receipts; a knot that is not admitted is not routed.

House rule: if used in prod, R2; if not, Hugging Face. R2 carries the serving surface (bytes a live lane can fetch). Experimental and retired knots belong here so the serving bucket stays small and auditable.

Inventory

artifact bytes notes
muse-glimmer-30b.knot 16,743,627,833 experimental, not in prod
shards-glimmer-3/shard-{0-entry,1-mid,2-exit}.knot 7,129,089,937 / 4,258,208,853 / 5,356,329,205 layer-sharded glimmer set (derived)
shards-glimmer/shard-0-entry.knot 9,684,015,101 glimmer entry shard (derived)
gemma4-31b-it-heretic-ara.knot 2,207,596,544 experimental, not in prod
falcon-mamba-7b-instruct.knot 8,157,731,597 experimental, not in prod
mistral-7b.knot 7,701,832,989 experimental, not in prod
qwen2.5-7b-instruct-7b.knot 4,677,153,133 Qwen2.5 7B variant
qwen2.5-7b-instruct-q4_k_m.knot 4,677,153,145 Qwen2.5 7B Q4_K_M variant
idol-turbo.knot 3,638,843,041 evaluation knot
qwen2.5-0.5b.knot 669,793,401 small-model checkpoint knot
qwen2.5-1.5b-q4km.knot 980,129,997 small-model Q4_K_M knot

Uploaded from LAN hosts over 2026-11-13; every file was SHA-256 hashed before transport and its byte length verified against the HF tree after landing.

Production artifacts (R2)

The admitted G1L ladder β€” rwkv7-g1l-1.5b, 2.9b, 7.2b, 13.3b β€” lives in R2 and is mirrored as model cards at forkjoin-ai/rwkv7-g1l-{1.5b,2.9b,7.2b,13.3b}. Those cards are the authority for provenance, quant policy, and admission: the Paris gate [6699,51128,4706,44312,4600] -> 37138 must pass before a knot can be routed.

Serving surface: apps/edgework-app

Edgework (apps/edgework-app) is the Cloudflare-facing app shell, API surface, browser transport, and proof-backed Aeon delivery surface that fronts inference as a priced workload. It admits compute/replay requests and issues settlement receipts; it is not a model and carries no weights. The knots in this organization are served behind that surface through the admission gate, never directly from this archive. A card-only repository for the surface lives at forkjoin-ai/edgework-app.

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. πŸ™‹ Ask for provider support