Capital Markets AI Training Data — Platinum Tier

643,382 Verified CRE Pairs
Across 10 Specialties

Underwriting calculations, IC memos, lease intelligence, debt sizing, market comparables, and risk triage. Every pair quality-gated, SHA-256 sealed, and ready for fine-tuning.

Request Access → Talk to Morey →
643,382 Verified Pairs
10 Specialties
19 SwarmSkills Live
Platinum Quality Tier

Deal Analytics & Underwriting

206,000+ pairs
Underwriting Calculations PLATINUM
206,282 pairs Cap rate analysis, NOI calculation, DSCR sizing, yield-on-cost, IRR projection, and going-in/exit cap modeling. The foundation of CRE deal math.
Risk Triage PLATINUM
32,169 pairs Red flag detection, deal screening, environmental risk, tenant concentration risk, and market timing assessment. Binary PASS/FAIL with rationale.

Investment Committee Intelligence

141,000+ pairs
IC Memo Drafting PLATINUM
141,544 pairs Full investment committee memo generation. Deal thesis, financial summary, risk analysis, market positioning, and recommendation. Institutional-grade output.

Lease Intelligence

141,000+ pairs
Lease Reasoning PLATINUM
96,507 pairs Lease structure analysis, tenant retention strategy, market rent comparison, escalation modeling, and renewal probability assessment.
Lease Abstraction PLATINUM
45,034 pairs Automated key terms extraction from commercial leases. Base rent, CAM, TI, free rent, renewal options, co-tenancy, and exclusivity clauses.

Market Comparables & Analysis

70,000+ pairs
Market Comparables PLATINUM
70,772 pairs Comparable sales and lease analysis. Price per square foot, cap rate benchmarking, submarket positioning, and adjustment methodology.

Financial Operations

90,000+ pairs
T12 Analysis PLATINUM
45,034 pairs Trailing 12-month P&L analysis. Revenue trending, expense ratio benchmarking, NOI bridge, and operational efficiency assessment.
Rent Roll Analysis PLATINUM
45,034 pairs Rent roll ingestion and intelligence. Occupancy analysis, lease expiration scheduling, mark-to-market, and revenue projection modeling.

9 Asset Types

Full industrial coverage

Every pair is tagged with its asset type. The dataset covers the full spectrum of industrial and commercial real estate, from last-mile logistics to GPU-grade data centers.

infill warehouse small bay flex industrial cross dock cold storage IOS (industrial outdoor) micro fulfillment data center industrial land

19 SwarmSkills — Live API

Every skill is a POST endpoint on router.swarmandbee.ai. Schema-validated, mock-tested, stored to R2.

broker_senior
Senior broker deal analysis. Pricing, risk scoring, deal structure, and go/no-go recommendation.
POST /skill/broker_senior
broker_junior
Junior broker research. Comparable analysis, market context, and property profiling.
POST /skill/broker_junior
intelligence_query
CRE database and market intelligence queries with structured results.
POST /skill/intelligence_query
debt_analyzer
Loan sizing, DSCR, LTV, refinancing analysis, and capital structure optimization.
POST /skill/debt_analyzer
bookmaker
Deal bookmaking and financial structuring for investment offerings.
POST /skill/bookmaker
deal_tracker
Active deal tracking, pipeline management, and status updates.
POST /skill/deal_tracker
comp_analyzer
Comparable property analysis with adjustments and market positioning.
POST /skill/comp_analyzer
rent_roll_analyzer
Rent roll ingestion, occupancy analysis, and revenue optimization.
POST /skill/rent_roll_analyzer
exchange_1031
1031 exchange analysis, timeline guidance, and replacement property identification.
POST /skill/exchange_1031
portfolio_optimizer
Portfolio strategy, diversification analysis, and rebalancing recommendations.
POST /skill/portfolio_optimizer
developer
Developer/investor feasibility analysis for ground-up and value-add projects.
POST /skill/developer
investor
Investor profiling, matching, and LP/GP structuring analysis.
POST /skill/investor
market_report
Market report generation with submarket trends, vacancy, and rent growth.
POST /skill/market_report
site_selector
Site selection and ranking based on demographics, access, and market factors.
POST /skill/site_selector
tax_assessor
Tax assessment analysis, appeal viability, and assessed value benchmarking.
POST /skill/tax_assessor
lead_scorer
Lead scoring and prioritization based on deal probability and fit.
POST /skill/lead_scorer
signal_scraper
Market signal detection from listings, filings, and public records.
POST /skill/signal_scraper
email_composer
Deal-related email composition for outreach, follow-up, and LOIs.
POST /skill/email_composer
news_digest
CRE news digest and market summary from multiple intelligence sources.
POST /skill/news_digest
Example API Call
curl -X POST https://router.swarmandbee.ai/skill/debt_analyzer \
  -H "Content-Type: application/json" \
  -d '{"property_type":"multifamily","noi":2100000,"loan_amount":25000000,"rate":0.068}'

Trained Models

Production models fine-tuned on SwarmCapitalMarkets data. From 0.8B edge to 122B institutional.

SwarmCRE-122B "Founder FTW"
122B params — fp8, 2 GPUs

Full-stack CRE reasoning and portfolio strategy. The institutional-grade model for complex deal analysis, multi-property portfolios, and strategic investment decisions.

founder tier 122B params fp8 precision 2x GPU
SwarmCRE-9B "Morey"
Qwen3.5-9B — Mamba-Transformer hybrid

The flagship CRE analyst. 9.5B parameters trained on 643K CRE intelligence pairs. Underwriting, IC memos, deal analysis, debt sizing, rent roll analysis, market intelligence. Chat and voice interface.

sealed 643K CRE pairs chat + voice 49.4 tok/s
SwarmCRE-35B
Qwen3.5-35B-A3B — bf16 LoRA r=64

Deep CRE specialist. Mixture-of-experts architecture with 3B active parameters. Extended reasoning for complex underwriting and multi-asset portfolio analysis.

sealed v1 87,940 pairs MoE 3B active GGUF Q4_K_M
SwarmCRE-4B
Compact edge model

Edge deployment target. Fits Jetson Orin Nano and BeeBox appliance. Local CRE intelligence without cloud dependency.

4B params edge-ready Jetson / BeeBox
SwarmCRE-2B
Ultra-light CRE analyst

Desktop and mobile deployment. 2GB VRAM. Fast local inference for deal screening and quick valuations.

2B params 2GB VRAM desktop + mobile
BeeMini-3B
Qwen2.5-3B-Instruct — Skill Router

Routes incoming queries to the correct SwarmSkill. 98.3% valid JSON output. 1.8GB Q4_K_M. The traffic controller for the entire skill economy.

sealed v2 98.3% JSON 60K pairs 1.8GB GGUF

Delivery

Every order ships in 5 formats with train/eval splits, provenance docs, and tamper-evident verification.

5
Output Formats
95/5
Train / Eval Split
SHA-256
Sealed Guarantee
DATA_CARD
Full Provenance
5 Delivery Formats — All Include Train + Eval Split
1. ChatML — swarmcm_train.chatml.jsonl
OpenAI API, TRL, Unsloth, Axolotl
2. Alpaca — swarmcm_train.alpaca.jsonl
LLaMA-Factory, HuggingFace trainers
3. ShareGPT — swarmcm_train.sharegpt.jsonl
FastChat, Vicuna, multi-turn
4. OpenAI — swarmcm_train.openai.jsonl
Direct gpt-4o fine-tuning upload
5. Completion — swarmcm_train.completion.jsonl
Legacy pipelines, custom loops
DATA_CARD.json
Quality metrics, model lineage, gate pass rates, specialty distribution, generation model ID.
guarantee.json
Merkle root of every pair. SHA-256 sealed. Optional Hedera HCS on-chain timestamp. Tamper-evident.
6 Gates
Deterministic quality: JSON validity, output length, numeric verify, concept presence, dedup, degeneration.
Per-Pair
Every pair carries: source, order_id, specialty, asset_type, model, quality gate result, content fingerprint.