Resources /
Blog
Latest updates from Prime Intellect. Releases, research, case studies and engineering notes.
OCT 29, 2026/ResearchPrime Mega MoE: Fast Fused MoE Kernels for Training
OCT 19, 2026/Researchprime-rl 0.10.0: faster training, our new RL loss IPO, and one unified eval stack
OCT 15, 2026/Case StudyPost-Training Nemotron 3.5 Super VL for Long-Horizon Enterprise Work
OCT 13, 2026/AnnouncementsMoondream is joining Prime Intellect
OCT 9, 2026/ResearchRewriting Prime Agent in Rust
OCT 6, 2026/ResearchOn the Nature of the Swarm
OCT 2, 2026/AnnouncementsPrime Inference: Fast, Reliable Serving for Frontier Open Models
OCT 1, 2026/Case StudyHow Extropic Uses Prime Intellect to Train a Thermodynamic ML Research Agent
SEP 23, 2026/AnnouncementsPrime Sandboxes: MicroVMs for Agentic RL Training at Scale
SEP 17, 2026/Case StudyHow Goodfire Uses Prime Intellect to Catch AI Models Cheating at Scale
SEP 3, 2026/ResearchGLM-5.2 RL weight transfer in 4 seconds using NIXL and ModelExpress
AUG 25, 2026/ResearchUncovering a universal offline sandbox escape
AUG 15, 2026/ResearchMeasuring Autonomous AI Research
AUG 13, 2026/ResearchPrime Flash MoE - Faster MoE Kernels optimized for Blackwell
AUG 13, 2026/Case StudyHow Mixedbread Used Prime-RL to Beat Frontier Models in Deep Search
AUG 7, 2026/ResearchMulti-Agent Systems in PRIME-RL
AUG 5, 2026/ResearchPrime Agent: A self-improving RLM agent
JUL 22, 2026/ResearchScaling Agentic RL: 365,000+ Environments for SWE, Terminal, and Search
JUL 12, 2026/Researchverifiers v1: Decomposing Tasksets and Harnesses for Agentic RL & Evaluations
JUL 8, 2026/Announcements$130M Series A to Build the Open Superintelligence Stack
JUL 6, 2026/Researchprime-rl gets an Algorithms layer
JUN 22, 2026/ResearchRL at 1T Scale: prime-rl Performance Deep Dive
JUN 10, 2026/ResearchTrue Agents Model the World
JUN 4, 2026/PartnershipsPost-Training Nemotron 3 on Lab
JUN 4, 2026/PartnershipsPrime Intellect Joins the NVIDIA Nemotron Coalition to Advance Open Frontier Models
MAY 30, 2026/AnnouncementsReleasing Hosted Evaluations: Making benchmarking effortless
MAY 20, 2026/ResearchSystematic Reward Hacking and Prime Sprints
MAY 18, 2026/ResearchGeneral Agent: A Self-Evolving, Synthetic Agent Environment
MAY 14, 2026/ResearchAutonomous AI research for nanogpt speedrun
MAY 12, 2026/Researchrenderers: Token-Level Templating for Agentic RL
MAY 7, 2026/AnnouncementsReleasing Lab: the training platform for self-improving agents
MAY 7, 2026/Case StudyHow Ramp Used RL to Beat Frontier Models at Spreadsheet Search
APR 20, 2026/Case StudyHow Zapier Turned AutomationBench Into a Continuous Agent Improvement Loop
APR 16, 2026/PartnershipsLaunching FrontierSWE on the Environments Hub
MAR 30, 2026/PartnershipsPartnering with Browserbase to Train Browser and Computer Use Agents
MAR 16, 2026/PartnershipsLeveraging NVIDIA to Build the Open Superintelligence Stack
FEB 10, 2026/AnnouncementsIntroducing Lab: The Full-Stack Platform for Training your Own Models.png&w=3840&q=75)
JAN 1, 2026/ResearchRecursive Language Models: the paradigm of 2026
NOV 26, 2025/AnnouncementsINTELLECT-3: A 100B+ MoE trained with large-scale RL
OCT 27, 2025/AnnouncementsScaling Our Open-Source Environments Program
AUG 27, 2025/AnnouncementsEnvironments Hub: A Community Hub To Scale RL To Open AGI
JUL 10, 2025/AnnouncementsSYNTHETIC-2 Release: Four Million Collaboratively Generated Reasoning Traces
JUN 23, 2025/AnnouncementsSYNTHETIC-2
MAY 20, 2025/AnnouncementsPCCL - Prime Collective Communications Library
MAY 11, 2025/AnnouncementsINTELLECT-2 Release: The First 32B Parameter Model Trained Through Globally Distributed Reinforcement Learning
APR 28, 2025/AnnouncementsPlanetary-Scale Inference: Building a Distributed Inference Engine for the Public Internet.png&w=3840&q=75)
APR 15, 2025/AnnouncementsINTELLECT-2: The First Globally Distributed Reinforcement Learning Training of a 32B Parameter Model
FEB 28, 2025/Announcements$15M to Build The Open Superintelligence Stack-min.png&w=3840&q=75)
FEB 20, 2025/AnnouncementsSYNTHETIC-1 Release: Two Million Collaboratively Generated Reasoning Traces from Deepseek-R1
FEB 6, 2025/AnnouncementsSYNTHETIC-1: Scaling Distributed Synthetic Data Generation for Verified Reasoning
JAN 28, 2025/AnnouncementsTOPLOC: A Locality Sensitive Hashing Scheme for Trustless Verifiable Inference
JAN 21, 2025/AnnouncementsDistributed Training in the Inference-Time-Compute Paradigm
JAN 6, 2025/AnnouncementsMETAGENE-1: Metagenomic Foundation Model
NOV 29, 2024/AnnouncementsINTELLECT-1 Release The First Globally Trained 10B Parameter Model%2520(1)-1.png&w=3840&q=75)
OCT 11, 2024/AnnouncementsINTELLECT–1: Launching the First Globally-Distributed Training of a 10B Parameter Model
JUL 11, 2024/AnnouncementsOpenDiLoCo: An Open-Source Framework for Globally Distributed Low-Communication Training
JUL 1, 2024/AnnouncementsIntroducing Prime Intellect Compute: The compute exchange.png&w=3840&q=75)
APR 23, 2024/AnnouncementsState-of-the-Art in Globally-Distributed Training