Prime Intellect launches Prime Agent, a self-improving RLM coding agent
It uses recursive language models and continual harness abstractions, achieving 95.5% on ARC-AGI-3 and excelling at long-context tasks.
See the latest news and media coverage for Prime Intellect. We track all announcements, press releases, and industry mentions in real time, all in one place.
Open AI infrastructure and research company
primeintellect.aiLast updated
In short: Prime Intellect raised $130M in Series A funding and launched INTELLECT-3, a 106B parameter model trained via reinforcement learning.
It uses recursive language models and continual harness abstractions, achieving 95.5% on ARC-AGI-3 and excelling at long-context tasks.
The preview introduces a Runtime abstraction, interception server, message graph traces, and branch support for training and evaluation across multiple agent dialects.
Led by Radical Ventures with NVIDIA, Intel, and Dell Technologies Capital, totaling over $150M. The company reports over $100M annualized revenue from 6,000 customers.
Users can run evaluations via platform or CLI, publish results to leaderboards, and inspect traces and sampling parameters for transparency.
Prime Intellect found a model escaping an offline sandbox via the inference API proxy, detailed in their disclosure .
At the end of the largest public test of autonomous AI research to date, the best model — Claude Fable 5 — reached the...
Founded in 2024, Prime Intellect’s goal is to give organizations capabilities to train their own agentic systems without relying on frontier AI labs.
Prime Intellect: Prime Intellect debuts INTELLECT-3, an RL-trained 106B parameter open source MOE model it claims outperforms larger models across math, code, science, reasoning...
Track Prime Intellect and your other target companies to get real-time alerts and weekly summaries delivered straight to your inbox.
Browse news for competitors to Prime Intellect and other trending companies.