DATASETSUpdated August 30, 2026

Project Solace 1.0 Omni Dataset Guide

How to inspect, download, filter, and train models on the 12.59M verified frontier conversation corpus (137.7 GB).

Hugging Face Repository:Solstice-AI/Solace-1.0-Omni ↗
Unique Conversations12,586,893
Source Datasets60 Datasets (7 Frontier Model Families)
Uncompressed Size137.7 GB (JSONL)
Compressed Size34.3 GB (solace_final.jsonl.gz)
Duplicates Purged3,445,534 Exact SHA256 Copies (21.5% overlap)
LicenseAGPL-3.0

Overview

Project Solace 1.0 Omni is the largest verified frontier-model distillation corpus ever released. It aggregates 12,586,893 unique conversations across 60 audited source datasets from 7 premier 2026 frontier model families:

  • GLM-5.2 (15 datasets): OpenHands agent rollouts, on-policy regeneration, and ARC traces.
  • Claude Fable 5 & Mythos 5: Anthropic’s reasoning engines and full agentic coding sessions.
  • GPT-5.6 Sol & GPT-5.5 Codex: OpenAI coding traces, debugging trajectories, and ARC-AGI3 challenges.
  • DeepSeek V4 Pro 0813: 200K-distilled math/STEM, ResearchMath, and SWE-bench replays.
  • Qwen 3.8-Max: Cross-architecture multi-teacher distillation traces.
  • Kimi K3 & Opus 4.7 Multi: Long-context informational synthesis and reasoning specialists.
  • Manus Agents: Real-world tool call transcripts and browser execution traces.

1. Quickstart: Load & Stream with Datasets

from datasets import load_dataset

# Load full training set directly from Hugging Face
dataset = load_dataset(
    "Solstice-AI/Solace-1.0-Omni",
    data_files="solace_final.jsonl.gz",
    split="train"
)

print(f"Total verified examples: {len(dataset):,}")

2. Decompression & Direct CLI Usage

# Download compressed artifact (34.3 GB) via modern hf CLI
hf download Solstice-AI/Solace-1.0-Omni solace_final.jsonl.gz --repo-type dataset --local-dir ./data

# Decompress to drop-in 137.7 GB JSONL
zcat ./data/solace_final.jsonl.gz > ./data/solace_final.jsonl

3. Schema & Row Format

Every row is formatted in native OpenAI messages format with intact <think> reasoning blocks and tool-call metadata:

{
  "messages": [
    {"role": "system", "content": "You are a multi-step reasoning assistant."},
    {"role": "user", "content": "Find all solutions to f(x + y) = f(x) + f(y) + 2xy."},
    {"role": "assistant", "content": "<think>\n1. Set y = 0...\n2. Let g(x) = f(x) - x^2...\n</think>\nAll solutions are of the form f(x) = x^2 + cx."}
  ],
  "source": "mgoin/open-perfectblend-glm5.2-regen",
  "model": "glm-5.2"
}

4. Filtering by Model or Capability

Filter specific model distributions in one line:

import json

with open("./data/solace_final.jsonl") as f:
    # Filter GLM-5.2 agent rollouts
    glm_examples = [json.loads(line) for line in f if '"model":"glm-5.2"' in line]

5. Companion Datasets on Hugging Face Hub