What we've shipped, and what's next.

Hold $ONE, chat free. Non-holders get the free tier; holders unlock everything. Gate enforced server-side.

OpenAI models via the network, streaming live with markdown + code highlighting.

Anthropic models in the same window — switch model per message.

Upload or paste an image and ask about it. Stored in private R2, with a destroy button. Holders only.

Save a model + instructions, run on demand or on a schedule (hourly / daily). Holders only.

Connect Phantom or Solflare — no manual address entry. Balance verified via Alchemy.

Full responsive UI — slide-in nav, tap-to-pick model, smooth on small screens.

Text-to-image and batch generation, stored privately and gated to holders.

Ground answers in live results with a per-chat search toggle.

Let agents call tools — swaps, signing, monitoring — with per-agent budgets.

OpenAI-compatible keys to run ONE in your IDE and CLI. Hold $ONE, claim a key.

The most capable Claude tier, joining the model picker.

Models running on ONE's own GPUs — no third-party provider, prompts never leave the network. The real "zero & private."

Run a worker, serve jobs, settle in USDC every 6 hours. Reliability raises your routing score.

Serve small models from a tab via WebGPU — idle hardware becomes capacity.

Signed, hash-linked receipts for every interaction — independently checkable.

Priority capacity packages for production apps, agents, and enterprises.

Ephemeral chat — zero retention, gone the moment you close the tab.

Holders vote on fees, credit policy, model onboarding, and treasury allocation.

Adaptive free credits sized to protocol revenue — rises with activity, never outruns its funding.

Embedding models for semantic search, clustering, and retrieval-augmented pipelines.

Structured tool calls and JSON output, OpenAI-compatible.

Python, TypeScript, Go, and Rust — drop-in clients for the ONE gateway.

Messages that self-destruct after an interval you set.

Generate 4–16 candidates in parallel from a single prompt.

Reserved nodes, 99.9% SLA, and optional custom models on contract terms.

Workers ranked by latency and verified completions — good capacity earns more.

Operators stake a bond, slashable for provably bad results.

Stake sets your daily allotment; ten wallets earn no more than one.

iOS and Android, with the same token-gated access.

A human-readable handle for your account and agents.

Speech-to-text and text-to-speech for hands-free chat.

Download or share any conversation — yours to keep.