Downloads · 30 days
7
23% of all-time downloads
metafloor-ai/manifest-orchestrator-9b-v0.6.0
manifest-orchestrator-9b-v0.6.0 is a text generation model from metafloor-ai. Use it when you need the model to write or continue text. It is set up for peft. The card lists the license as cc-by-nc-4.0.
Preferred 60.4% of the time over the base model — 66.0% with matched format On 134 held-out expert questions, an independent two-model LLM judge panel preferred Manifest 9B's answer over the untuned base model's answe…
Downloads · 30 days
7
23% of all-time downloads
All-time downloads
31
Public
Repo size
136 MB
Likes
0
Public
Click a slice to open those files.
.safetensors116 MB · 85%
From the Hugging Face model README
Preferred 60.4% of the time over the base model — 66.0% with matched format
On 134 held-out expert questions, an independent two-model LLM judge panel preferred Manifest 9B's answer over the untuned base model's answer 60.4% of the time (95% CI 55.6–65.3%; 37 wins / 9 losses / 88 ties). With both models given the same answer format, it is preferred 66.0% of the time (95% CI 61.2–70.9%; 50 / 7 / 77). Both clear the ship bar (CI lower bound > 55%) — specialization confirmed.
Win-rate is measured against each model's own base. The 9B base is already very strong, so many answers tie — the margin is narrower than the smaller Manifest models, but Manifest 9B still wins outright and adds real domain depth on the hardest questions.
Manifest is MetaFloor's suite of supply-chain expert models — purpose-built specialists in procurement, demand planning, warehouse operations, supplier relationship management, risk & resilience, transportation, inventory, and order fulfilment.
The hardest, most nuanced supply-chain questions — multi-constraint trade-offs, cross-functional reasoning, and long, structured answers — where the extra capacity of a 9B base pays off. For lighter workloads or tighter latency/cost budgets, the 2B and 4B orchestrators are strong, cheaper options.
Manifest 9B is a LoRA adapter — load it on top of its base model:
from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel
base = "Qwen/Qwen3.5-9B"
tok = AutoTokenizer.from_pretrained(base)
model = AutoModelForCausalLM.from_pretrained(base, device_map="auto")
model = PeftModel.from_pretrained(model, "metafloor-ai/manifest-orchestrator-9b-v0.6.0")
msgs = [{"role": "user", "content": "How should I set safety stock across a 3-echelon network with 6-week supplier lead times?"}]
inputs = tok.apply_chat_template(msgs, add_generation_prompt=True, return_tensors="pt").to(model.device)
print(tok.decode(model.generate(inputs, max_new_tokens=512)[0][inputs.shape[1]:], skip_special_tokens=True))
Prompt tip: Manifest is trained to condition on the scenario — include the asker's role and operating scale (revenue, SKUs, suppliers, nodes, lead time) in the message for the sharpest, most tailored answers.
Two kinds of models:
🧭 Orchestrators — general-purpose, handle any supply-chain area
| Model | Size | Preferred over base | Status |
|---|---|---|---|
| Manifest 0.8B | 0.8B | 79.9% | ✅ available |
| Manifest 2B | 2B | 85.1% | ✅ available |
| Manifest 4B | 4B | 95.1% | ✅ available |
| Manifest 9B | 9B | 60.4% | ✅ available |
🎯 Domain-experts — specialized for a single area
| Model | Preferred over base | Status |
|---|---|---|
| Manifest Specialist · Risk & Resilience | 100% (20 / 0) | ✅ available |
| Manifest Specialist · Inventory Optimization | 80.0% | ✅ available |
| Manifest Specialist · Demand Planning | 82.5% | ✅ available |
Orchestrators are scored on the general supply-chain benchmark; domain-experts on their focused domain benchmark. Win-rate is measured against each model's own base, so figures are not directly comparable across sizes — a stronger base is a tougher opponent.
134 held-out expert questions across 8 supply-chain areas. Each question is answered by Manifest and by the base model (given the same answer format); an independent two-model LLM judge panel then picks the better answer. Manifest 9B was preferred 66.0% of the time with matched format (95% CI 61.2–70.9%; 50 wins / 7 losses / 77 ties over 134). Benchmark: supply-chain-eval.
| Method | LoRA (PEFT), rank 16 / alpha 16 / dropout 0.05 |
| Target modules | all attention + MLP projections |
| Trainable params | 29,097,984 (~29.1M) |
| Epochs | 2 |
| Training examples | ~32,000 |
| Final loss | 1.24 |
Training data: supply-chain instruction–response pairs from the seed-driven operator-as-teacher pipeline — a deterministic engine emits a unique seed per example (area, sub-area, persona, question type, realistic numeric scenario) and a strong teacher model writes the matching answer. The training data is drawn from MetaFloor's proprietary ~32k-example supply-chain dataset, which is not open-sourced — only the held-out evaluation benchmark (supply-chain-eval) is public.
Manifest models and the supply-chain-eval benchmark are released under CC-BY-NC-4.0 — free for research and non-commercial use, with attribution. Commercial use requires a license from MetaFloor — get in touch at metafloor.ai.
Manifest 9B is a LoRA adapter over Qwen/Qwen3.5-9B (used under its own license); the base model is required to load the adapter.
<!-- manifest-suite:order -->