← All companies

Frontier / Official source checked

Prime Intellect

What Prime Intellect does

Prime Intellect is a private AI “open superintelligence” lab and platform company building an end-to-end stack for training, evaluating, deploying, and continuously improving agentic models—while emphasizing globally sourced compute and open research. The company’s product surface describes a unified loop across RL environments, hosted evaluations, training, inference, and sandboxed execution, positioned so teams can specialize open models for their own workflows rather than relying only on closed frontier APIs. Prime Intellect’s technical approach centers on reinforcement-learning (RL) and “agent harness” abstractions intended to turn tasks into verifiable RL environments, run training at scale, benchmark improvements, and then deploy the resulting models with continued evaluation.

News

Company record

Aug 14, 2026 · Official · Prime Intellect Blog (Releases feed)Measuring Autonomous AI Research (Prime Intellect release)

Prime Intellect published details about autonomous AI research measurement and reporting it as a release item on its blog feed for 2026.

Aug 05, 2026 · Official · Prime Intellect BlogPrime Agent: a self-improving RLM agent launch

Prime Intellect launched Prime Agent, describing a self-improving coding harness built around Recursive Language Models (RLM) and a Continual Harness, intended to support adaptation of agent behavior while running.

Jul 08, 2026 · Official · Prime Intellect BlogPrime Intellect raises $130M Series A to build the Open Superintelligence Stack

Prime Intellect announced a $130M Series A led by Radical Ventures, with participation including NVIDIA Ventures, Intel Capital, and Dell Technologies Capital, as it scales compute, RL, evaluations, inference, and continual learning.

Jul 05, 2026 · Official · Prime Intellect BlogPrime Intellect releases/verifies prime-rl “Algorithms layer” for RL credit assignment

Prime Intellect published an update describing an algorithms-layer component for prime-rl focused on credit assignment from rollouts to training signals (e.g., rewards/advantages), and plans for making the pipeline more expressive.

Jun 04, 2026 · Official · Prime Intellect BlogPrime Intellect joins the NVIDIA Nemotron Coalition to advance open frontier models

Prime Intellect announced it joined the NVIDIA Nemotron Coalition and described its contribution as focused on post-training/reinforcement learning environments, verifiers, Prime Sandbox, and NeMo Gym integration, with an emphasis on continuous improvement loops.

May 28, 2026 · Official · Prime Intellect Blog (Releases feed)Releasing Hosted Evaluations: making benchmarking effortless

Prime Intellect announced hosted evaluations intended to simplify benchmarking for teams building and improving models within its stack.

Mar 30, 2026 · Official · Prime Intellect BlogPartnering with Browserbase to train browser and computer use agents

Prime Intellect announced a partnership with Browserbase, describing BrowserEnv as a training experience for browser/computer-use agents, and a proof via fine-tuning Qwen3-VL-8B-Instruct on WebVoyager-based live web tasks.

Mar 16, 2026 · Official · Prime Intellect BlogLeveraging NVIDIA to build the open superintelligence stack

Prime Intellect detailed how its collaboration with NVIDIA supports its stack across training and serving, including mention of NVIDIA Blackwell hardware and NVIDIA Dynamo for inference orchestration, and related sandboxing and optimization components.

Show 1 earlier updates
Source map · 5 recurring channels · 23 references

Funding

Latest disclosed valuation$1B
Tracked capital$150.5M3 sourced rounds

What it builds

Prime Intellect Lab

Lab is Prime Intellect’s training platform for turning ideas into better models by integrating environments, training, and evaluation as part of the broader open superintelligence stack.

source ↗
Prime Intellect Compute Exchange

Compute Exchange is Prime Intellect’s compute marketplace intended to connect GPU supply and demand and support training and related infrastructure workflows.

source ↗
Prime Intellect Lab (hosted evaluations + loop)

Prime Intellect describes an integrated loop spanning RL environments, hosted evaluations (benchmarking), training, inference deployment, and sandboxed execution.

source ↗
Prime Agent

Prime Agent is a “self-improving RLM agent” launched by Prime Intellect as a coding harness built around Recursive Language Models (RLM) and a Continual Harness.

source ↗
prime-rl and verifiers (open-source libraries)

Prime Intellect’s open-source libraries include prime-rl (asynchronous RL at scale) and verifiers (used for building/verifying agentic RL environments and evaluations), which the company says integrate into the Lab workflow.

source ↗
Environments Hub (RL environment catalog)

The Environments Hub is Prime Intellect’s community hub for RL environments, positioned as a growing collection of environments for training and evaluation.

source ↗

Milestones & partnerships

Aug 05, 2026
Prime Agent launch

Prime Intellect launched Prime Agent, a self-improving RLM agent harness described as built around Recursive Language Models and a Continual Harness.

Jul 08, 2026
Series A announced for scaling the full stack

Prime Intellect announced a $130M Series A led by Radical Ventures to scale compute clusters, RL runs, evaluations, inference, and continual learning.

Radical Ventures · NVIDIA Ventures · Intel Capital · Dell Technologies Capital
Jul 05, 2026
prime-rl “Algorithms layer” update

Prime Intellect published details of an algorithms layer focused on producing final rollout signals for credit assignment and planned expansions (sampling, judging, credit assignment).

Jun 04, 2026
NVIDIA Nemotron Coalition membership

Prime Intellect announced it joined the NVIDIA Nemotron Coalition with contributions focused on post-training RL environments and tooling.

NVIDIA · Black Forest Labs · Cursor · LangChain · Mistral AI · Perplexity · Reflection AI · Sarvam · Naver · H Company · Nous Research · Thinking Machines Lab
Mar 30, 2026
Browserbase partnership for browser/computer-use agent training

Prime Intellect described partnering with Browserbase to support live web training for browser and computer use agents via BrowserEnv and related workflows.

Browserbase
Mar 16, 2026
NVIDIA collaboration spotlight

Prime Intellect described collaboration spanning training/serving and inference orchestration, plus integration into its stack.

NVIDIA
Feb 28, 2025
$15M funding announcement for the open superintelligence stack

Prime Intellect announced raising $15M led by Founders Fund to accelerate the stack across compute access, RL training, and open research.

Founders Fund · Menlo Ventures
Apr 23, 2024
Seed financing announced; Prime Intellect protocol/computing platform roadmap

Seed round announcement positioned around building the Prime Intellect protocol and compute platform enabling decentralized training and model co-ownership.

Distributed Global · CoinFund