In October 2024, a team from Google, Palantir, and Two Sigma walked into a16z's crypto accelerator with a thesis most of the room found uncomfortable: open-source AI is a lie. Not the models. There are 2.7 million of them on Hugging Face and anyone can download one. The lie is that downloading a model and running it are the same thing. They are not. Running a 70B-parameter LLM in production costs $30,000 in GPU hardware before the first electric bill. For most builders, the economics push you right back to the same three API providers who set the pricing, write the content policies, and decide which countries get access. The models were democratized. The infrastructure to run them was not. OpenGradient built the fix: a decentralized network where every AI inference is cryptographically verifiable without trusting any single party. Web2 speed. Onchain proof. Every model call auditable. The network has processed over 2 million verified inferences and generated 500,000+ cryptographic proofs. BitQuant, their DeFi agent, has served 1.8M users. The team raised $9.5M led by @a16zcrypto, backed by @cbventures, @SVAngel, and angel investors including @balaboris (former CTO, Coinbase), @ilblackdragon (Co-founder, NEAR), and @sandeepnailwal (founder, Polygon). Today we announce @OpenGradient will launch on Virtuals Protocol. 42% of all x402 transactions already run through Virtuals. 40,000+ agents tokenized. $13.5B in cumulative volume. The agent economy had a commerce layer. It did not have a verifiable intelligence layer. Now it does.

Introducing Eastworlds We help robots leave the lab faster Discover More eastworlds.io/