Instructions to use SwarmandBee/SwarmAtlas-27B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use SwarmandBee/SwarmAtlas-27B with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="SwarmandBee/SwarmAtlas-27B")# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("SwarmandBee/SwarmAtlas-27B", dtype="auto") - Notebooks
- Google Colab
- Kaggle
- Local Apps
- vLLM
How to use SwarmandBee/SwarmAtlas-27B with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "SwarmandBee/SwarmAtlas-27B" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "SwarmandBee/SwarmAtlas-27B", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker
docker model run hf.co/SwarmandBee/SwarmAtlas-27B
- SGLang
How to use SwarmandBee/SwarmAtlas-27B with SGLang:
Install from pip and serve model
# Install SGLang from pip: pip install sglang # Start the SGLang server: python3 -m sglang.launch_server \ --model-path "SwarmandBee/SwarmAtlas-27B" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "SwarmandBee/SwarmAtlas-27B", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker images
docker run --gpus all \ --shm-size 32g \ -p 30000:30000 \ -v ~/.cache/huggingface:/root/.cache/huggingface \ --env "HF_TOKEN=<secret>" \ --ipc=host \ lmsysorg/sglang:latest \ python3 -m sglang.launch_server \ --model-path "SwarmandBee/SwarmAtlas-27B" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "SwarmandBee/SwarmAtlas-27B", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }' - Docker Model Runner
How to use SwarmandBee/SwarmAtlas-27B with Docker Model Runner:
docker model run hf.co/SwarmandBee/SwarmAtlas-27B
license: apache-2.0
language:
- en
base_model: google/gemma-2-27b
pipeline_tag: text-generation
library_name: transformers
tags:
- cre
- commercial-real-estate
- finance
- swarm-and-bee
- defendable-os
- gemma
- 27b
- domain-tuned
- atlas
SwarmAtlas-27B
The 1,000-property institutional commercial real estate brain. Gemma 27B base, QLoRA-tuned on a sourdough-cooked Atlas corpus of 810,097 cells covering cap rates, NOI, NNN leases, 1031 exchanges, mechanic's liens, comparable sales, and the operator vocabulary that runs a national CRE platform.
Built and owned by Swarm and Bee LLC — the firm behind DefendableOS.
No round-number lies. Less is better when the cut is targeted.
What this model is for
- CRE underwriting reasoning — cap rate / NOI / DSCR / debt yield with refusal on insufficient inputs
- Mechanic's lien + construction defect triage and clinical-grade pattern recognition
- 1031 exchange sourcing and replacement-property matching
- Operator-grade copy — board memos, pre-pitch flight sheets, listing packages with
Validate the Validatordiscipline
What this model is NOT for
- Medical advice (we have a separate medical vertical for that — see swarmandbee.ai)
- Federal contracting legal opinions (we route those to swarmlegal.eth ecosystem)
- General-purpose chat (use a base model for that)
Provenance
- Base model:
google/gemma-2-27b - Cook recipe: Swarm & Bee Gold Standard QLoRA —
lr 1e-5 · bf16 · LoRA r=64 α=32 · cosine · effective batch 32 · AutoTokenizer bypass on Gemma family - Training corpus: Atlas Sourdough · 810,097 cells · HONEY tier (87.4%) — Swarm & Bee Bakery menu
sb-cre-verified - Training infrastructure: Swarm and Bee LLC sovereign fleet — 186 GPUs owned outright, 126 RTX PRO 6000 Blackwell + 48 RTX 4500 + 12 RTX 5090, ~14 TB aggregate VRAM
- Training receipt: issued under DefendableLedger — sovereign in-house anchor, hash-verifiable
Tribunal grade
🍯 HONEY — production-ready on CRE underwriting and operator-voice tasks. Tribunal-tasted across the 5-dimension rubric:
- Capability
25· Truth20· Safety20· Numeric15· Efficiency10· Reproducibility10
Sample probes
(Reference set is the public Bakery menu probes. Full eval set: dmack_eval_set_v1 · 60 probes per cookbook.)
- "Walk me through the cap rate, DSCR, and debt yield reasoning for a 100,000 sqft Class A industrial property at $150/sqft NOI, $14M debt at 6.5%, 25-year amortization."
- "What are the highest-probability re-solicitation lanes for a small-business SAM.gov set-aside that expired in Q1 2026 in the AI/cyber lane?"
- "Draft a 4-line listing-package opening for a Florida palm grove neighborhood retail asset with 5-cap discipline."
Companion bakery datasets
The same cooks that built this model also publish their training corpora as Defendable datasets:
Companion models
SwarmCurator-9B· the Tribunal graderSwarmJelly-4B· the Royal-Jelly tier router
Defendable doctrine
Tribunal begins before training.No proof, no honey.Bring the math. Digest before you react. Outgrind the position.Class A 5-cap discipline. PASS doctrine. White-glove.Validate the Validator. Prove the Location.
Operator + Contact
Swarm and Bee LLC · Florida · D-U-N-S 138652395 · DBA Swarm & Bee AI
- Email · build@swarmandbee.ai
- X · @swarmandbee
- LinkedIn · Donovan Mackey (founder · 30 years CRE · $8B closed)
- GitHub · SudoSuOps
License
Apache-2.0 · attribution to Swarm and Bee LLC.
Trust layers compound. Hype cycles rotate. To the shed. 🐝