BLACKBOX — Decentralized Inference Protocol

t token · b buy · f feed · d demo

BLACKBOX(1) Decentralized Inference Protocol BLACKBOX(1)

NAME

blackbox — permissionless AI inference routing on Solana

TOKEN

$BBOX·BBoX7VdR4GjLYb4ZfMTqNfAkGHUaHxkGE9VwL3pm4wEF

buy on jupiter dexscreener pump.fun

SYNOPSIS

bbox infer [--model model_id] [--route strategy] [--verify] prompt

DESCRIPTION

Blackbox routes AI inference requests across a network of compute providers. The protocol matches requesters with providers based on model availability, geographic proximity, and cost. Verification of inference results is handled via on-chain attestation.

The $BBOX token captures value through creator fees on every trade. Fees are split between automated buyback (80%) and an inference subsidy pool (20%) that reduces per-token costs for users.

ECONOMICS

Creator fees from all $BBOX trades on PumpSwap are routed to the protocol treasury automatically. Fee flow:

PumpSwap creator fees (every buy/sell)

├─ 80% → instant Jupiter buyback (continuous)

└─ 20% → inference subsidy pool

Every trade generates buy pressure. Buybacks execute continuously — no epochs, no delays. The inference pool subsidizes compute costs, creating a flywheel: more usage → more volume → more fees → more buyback.

TRY IT

Interactive inference demo. Try "help", "status", "fees", or any prompt.

blackbox — interactive inference● ready

blackbox inference engine v0.9.2-rc4

connected to routing mesh (2,841 nodes)

→ explain zero-knowledge proofs in one sentence

routing → us-east-1 (llama-70b)

A zero-knowledge proof lets you prove you computed something correctly without revealing the inputs or the process — Blackbox uses this to verify inference results without exposing model weights.

model: llama-70b latency: 18.9ms tokens: 142 cost: 0.001704 SOL

TELEMETRY

uptime14d 7h 23m 41sactive_nodes2,841inferences_1h847,291avg_latency_ms14.7p99_latency_ms41.2cache_hit_rate73.4%queue_depth12solana_slot400,431,812fees_captured0.00 SOLbuyback_modecontinuousupdated2026-02-15 15:18:20 UTC

NETWORK STATUS

us-east-1612 nodes8.3ms67% load

us-west-2441 nodes11.7ms54% load

eu-central-1528 nodes14.2ms71% load

ap-northeast-1387 nodes18.9ms43% load

sa-east-1203 nodes24.1ms89% load

me-south-1156 nodes19.4ms38% load

sa-east-1: elevated latency — investigating capacity constraints

INFERENCE FEED

Real-time inference requests across the network.

blackbox — inference feed (live)streaming

OPTIONS

--route optimal|nearest|cheapest

Select routing strategy. optimal balances latency and cost. nearest routes to the closest provider. cheapest minimizes per-token cost.

--model model_id

Target model. Supports Llama 70B, Mistral 7B, Gemma 2, DeepSeek V3, Codestral, Qwen 72B, and custom fine-tunes. Run bbox models --list.

--verify

Enable on-chain attestation of the inference result. Adds ~8ms overhead. Attestation hash is anchored to Solana for auditability.

--stream

Stream tokens as generated. Compatible with all text models. Attestation submitted after stream completion.

TOKEN DETAILS

ticker$BBOXchainSolanasupply1,000,000,000team_allocation0%tax0%lp_statusburnedlaunchpump.fun fair launch