A privacy-first multi-model AI inference network. 5 models live now, 55 across every major provider via OpenRouter — coming soon. Inference is memory-only, wiped on response, with a signed cryptographic receipt on every reply.
A small router (7–14B) classifies intent in under 3 seconds and selects the right model. In Broadcast and Diffuse modes all live models fire at once.
| Model | By | Type | Status |
|---|---|---|---|
| Claude Haiku 4.5 | Anthropic | text | live |
| Claude Sonnet 4.6 | Anthropic | text | live |
| Llama 3.3 70B | Meta | text | live |
| Kimi K2.7 Code | MoonshotAI | text | live |
| Kimi K2.6 | MoonshotAI | text | live |
| Claude Fable 5 | Anthropic | text | in build |
| Claude Opus 4.8 | Anthropic | text | in build |
| Claude Opus 4.8 Fast | Anthropic | text | in build |
| Claude Opus 4.7 | Anthropic | text | in build |
| Claude Opus 4.7 Fast | Anthropic | text | in build |
| Claude Opus 4.6 | Anthropic | text | in build |
| Claude Opus 4.6 Fast | Anthropic | text | in build |
| Claude Opus 4.5 Sonnet | Anthropic | text | in build |
| Llama 3.1 405B | Meta | text | in build |
| GPT-5.5 Pro | OpenAI | text | in build |
| GPT-5.5 | OpenAI | text | in build |
| GPT-5.4 Image 2 | OpenAI | image | in build |
| GPT Chat Latest | OpenAI | text | in build |
| Gemini 3.5 Flash | text | in build | |
| Gemini 3.1 Flash Lite | text | in build | |
| Gemini 3.1 Flash Image | image | in build | |
| Gemini 3 Pro Image | image | in build | |
| Grok 4.3 | xAI | text | in build |
| Grok Build 0.1 | xAI | text | in build |
| DeepSeek V4 Pro | DeepSeek | text | in build |
| DeepSeek V4 Flash | DeepSeek | text | in build |
| Qwen3.7 Max | Qwen | text | in build |
| Qwen3.7 Plus | Qwen | text | in build |
| Qwen3.6 Max Preview | Qwen | text | in build |
| Qwen3.6 Flash | Qwen | text | in build |
| Qwen3.6 35B A3B | Qwen | text | in build |
| Qwen3.6 27B | Qwen | text | in build |
| Qwen3.5 Plus | Qwen | text | in build |
| Mistral Medium 3.5 | Mistral AI | text | in build |
| Nemotron 3 Ultra | NVIDIA | text | in build |
| Nemotron 3 Nano Omni | NVIDIA | text | in build |
| North Mini Code | Cohere | text | in build |
| Granite 4.1 8B | IBM | text | in build |
| MiniMax M3 | MiniMax | text | in build |
| GLM 5.2 | Z.ai | text | in build |
| GLM 5.1 | Z.ai | text | in build |
| Ring-2.6-1T | inclusionAI | text | in build |
| Ling-2.6-1T | inclusionAI | text | in build |
| Ling-2.6 Flash | inclusionAI | text | in build |
| Hy3 Preview | Tencent | text | in build |
| MiMo-V2.5 Pro | Xiaomi | text | in build |
| MiMo-V2.5 | Xiaomi | text | in build |
| Laguna XS.2 | Poolside | text | in build |
| Laguna M.1 | Poolside | text | in build |
| Step 3.7 Flash | StepFun | text | in build |
| Perceptron Mk1 | Perceptron | text | in build |
| Nex-N2-Pro | Nex AGI | text | in build |
| Fusion | OpenRouter | text | in build |
| Owl Alpha | OpenRouter | text | in build |
| Pareto Code Router | OpenRouter | text | in build |
Inference pipeline: Router → Retrieval (optional) → Inference → Stream + Receipt → Wipe. Nothing is persisted.
Every response commits to its production conditions in a signed io_receipt_v1. Verifiable offline against the published Ed25519 public key — via SDK, MCP, CLI, or web.
Inspect source, watch the network tab, confirm no client-side storage, verify the signature binds response to model and policy. Misattribution and client-side leakage are detectable.
Workers run inside TEE-attested enclaves. The enclave_quote field binds the receipt to a measured binary in hardware-isolated environment. retained_prompt: false becomes hardware-backed, not operational.
Fair launch on PumpFun (Solana) — no pre-sale, no venture allocation. $IO coordinates access and payment across the network.
Inference paid in $IO via x402 micropayment protocol — discounted vs other assets.
Premium features — higher rate limits, broadcast/diffuse, proof certificates — gated by $IO.
Buyback-and-burn from net protocol revenue (after GPU & bandwidth) — supply tightens with usage.
At v1.0, holders vote on model registry, redaction policy, fee splits, and warrant canary cadence.
Fee policy note: Buyback-and-burn operates on net protocol revenue after infrastructure costs (GPU + bandwidth), not gross inference fees. The precise split between burn and operations reserve will be published as an on-chain parameter before activation.