#Toploc
Prime Intellect releases INTELLECT-2 (open-sourced first 32B parameter model trained via globally distributed reinforcement learning), which demonstrates that globally decentralized RL works, per them.

They also open sourced:
- PRIME-RL
- SHARDCAST
- TOPLOC Validators
May 12, 2025 at 6:11 AM
2/ Kern der Architektur: PRIME (dezentrales Training), GENESYS (synthetische Daten), TOPLOC (verifizierbare Inferenz) und ein starker Fokus auf post-training Reinforcement Learning. INTELLECT-2 und INTELLECT-3 sind erste Ergebnisse dieses Stacks.
July 8, 2026 at 4:47 PM
Jack Min Ong, Matthew Di Ferrante, Aaron Pazdera, Ryan Garner, Sami Jaghouar, Manveer Basra, Johannes Hagemann
TOPLOC: A Locality Sensitive Hashing Scheme for Trustless Verifiable Inference
https://arxiv.org/abs/2501.16007
January 28, 2025 at 5:48 AM
Prime has got to be one of the more underrated AI players.

They're not "sexy" in that they're not playing this game to dominate the AI space and make a trillion-bajillion dollars, but they're enabling fully decentralized AI that would be actually democratic and open.
TOPLOC - A Locality Sensitive Hashing Scheme for Trustless Verifiable Inference
We introduce TOPLOC, a novel method for verifiable inference. TOPLOC employs a compact locality sensitive hashing mechanism for intermediate activations, which can detect unauthorized modifications to...
www.primeintellect.ai
January 28, 2025 at 9:03 PM
PRIME-RL, our training framework purpose-built for distributed asynchronous reinforcement learning, based on top of novel components such as TOPLOC, which verifies rollouts from untrusted inference workers, and SHARDCAST, which efficiently broadcasts [3/6 of https://arxiv.org/abs/2505.07291v1]
May 13, 2025 at 6:21 AM