Pick a Pack — From $49
| Product | Rows | Tokens est. | Price | $/1k records | Buy |
|---|---|---|---|---|---|
Flagship Master Suite (Full 210k + 4 LoRAs) 🏆 FULL ENTERPRISE SUITE | 210,276 | ~880M+ Tokens | $249 | $1.18 | $249 |
Lab Starter — 2k Curated Samples 🚀 START HERE | 2,000 | ~8M+ Tokens | $49 | $24.50 | $49 |
Pack 3: Multi-Turn Agent Sandbox Episodes 🤖 AUTONOMOUS TOOL & COMPILER SESSIONS | 15,026 | ~70M+ Tokens | $79 | $5.26 | $79 |
Pack 2: Platinum Full-Stack Architecture 🌐 ENTERPRISE FULL-STACK & APIS | 102,508 | ~420M+ Tokens | $129 | $1.26 | $129 |
Pack 1: Systems & Concurrency Invariants ⚡ KERNEL & LOW-LEVEL CONCURRENCY | 92,742 | ~390M+ Tokens | $149 | $1.61 | $149 |
Flagship Master Suite (Full 210k + 4 LoRAs)
Unified archive including all 3 specialized domains, 4 pre-computed fine-tuned Gemma 4 LoRA adapters, AST invariant validation scripts, and hard SWE-bench evaluation harnesses.
Lab Starter — 2k Curated Samples
Curated starter trajectories spanning systems, fullstack, and agent domains so you can validate training pipelines before buying a full pack.
Pack 1: Systems & Concurrency Invariants
Specialized for training models that write rock-solid systems software in Rust, C, C++, Go, and Zig with memory safety, deadlock prevention, and cache-coherent algorithms.
Pack 2: Platinum Full-Stack Architecture
Engineered for full-stack coding agents handling complex TypeScript/Next.js, Python/FastAPI, GraphQL/REST APIs, database migrations, and cloud-native deployments.
Pack 3: Multi-Turn Agent Sandbox Episodes
Multi-turn episodes of agents navigating real shell environments, analyzing compiler diagnostics, recovering from broken test runs, and completing multi-step coding objectives.
Load into PyTorch / Hugging Face / Axolotl / Unsloth
Stream a purchased commercial archive into your fine-tuning pipeline.
from datasets import load_dataset
# Load complete downloaded archive into TRL / Axolotl:
ds_full = load_dataset(
"json",
data_files="payload_master_210k.tar.gz",
split="train"
)
print(f"Loaded {len(ds_full):,} System-2 records")Already purchased a dataset?
Enter your Stripe Checkout Session ID (starts with cs_live_ or cs_test_) to verify payment and unlock downloads in this browser.
Frequently Asked Questions
All records are delivered in standardized compressed JSONL (.jsonl in .tar.gz archives). Each row includes system directives, multi-turn tool conversations, and <thought> chain-of-thought blocks compatible with Unsloth, Axolotl, Llama-Factory, and Hugging Face TRL.
Purchased packages grant a perpetual, non-exclusive, worldwide license for internal model training, benchmarking, and commercial deployment and distribution of fine-tuned model weights with zero royalties.
Yes. For purchase orders (POs), corporate invoicing, or custom multi-node agreements, reach out directly to james@swarmagents.codes.
Once purchased, the license is bound to your account vault on SwarmAgents. You can return to this page anytime to stream direct downloads of your archives.