Docs
SUPA

Tokenized compute for super intelligence.

Launch a pump.fun bonding-curve asset bound to any SI model. Creator-fee flow is routed at the protocol level, 10,000 bps and immutable, into a deterministic per-mint compute vault that accrues as inference budget for that model. Every asset ships with an open terminal, on-chain verifiable routing.

Launch a coinExplore 418 modelsRead the docs
Frontier endpointsInput / output per 1M tokens
Claude Sonnet 5.5Anthropic$2.00 / $10GPT-6.1 Sol ProOpenAI$2.00 / $10Gemini 3.8 FlashGoogle$0.75 / $3.75Grok 4.7SpaceXAI$2.00 / $6.00DeepSeek V4.1 FlashDeepSeek$0.03 / $0.50Qwen3.8 Max PrimeQwen$4.00 / $12Llama 4 MaverickMeta$0.188 / $0.652Mistral Medium 3.5Mistral$1.50 / $7.50

Live deployments

No deployments indexed yet. The genesis slot is open.

Launch the first

Protocol flow

Full technical spec
  1. Model binding. Select an inference endpoint from 418 live models, each priced per million tokens.
  2. Atomic launch. One v0 transaction mints the bonding-curve asset on pump.fun and locks its fee-sharing config in the same slot.
  3. Fee-to-compute routing. 10,000 bps of creator fees accrue to a vault derived deterministically from the mint. Set once, immutable after.
  4. Open terminal. Try it before you launch. A free-tier preview engine serves replies during bootstrap; paired, token-metered inference activates per asset as compute accrues.