Prime Intellect's Verifiers v1 Rewrites AI Agent Training From the Ground Up
Prime Intellect overhauls its RL environment stack with composable tasksets, harnesses, and a smarter trace format that makes long-horizon agentic training actually feasible.
- Prime Intellect releases verifiers v1 (0.2.0), a complete rewrite of their RL environment stack for agentic training and evals.
- Environments are now decomposed into three composable pieces: a taskset (what), a harness (how), and a runtime (where).
- A new interception server proxies agent-to-model requests, enabling harness-agnostic trace recording and dialect normalization across OpenAI and Anthropic APIs.
- Traces are now stored as message DAGs, reducing storage complexity from O(n²) to O(n) in turns, making 100+ turn rollouts practical.
- Branching support lets compaction and subagent traces each produce independent training samples, enabling training past the model context window.
- GLM-4.5-Air was trained on ScaleSWE with 35-turn rollouts completing 1K steps in 2 days on 6 H200 nodes using v1 in production.
Training agents with reinforcement learning has a dirty secret: the environment layer is usually the bottleneck. Wiring up a benchmark, an agent harness, and a sandboxed runtime into something that can generate thousands of verified rollouts per day is painful, bespoke work. Prime Intellect's verifiers v1 is a ground-up rewrite of their environment stack designed to fix exactly that.
The Three-Part Split That Changes Everything
The central idea in v1 is a clean decomposition of what used to be a monolithic "environment" into three independent pieces:
- Taskset , defines the work: the data, tools, and scoring logic. It knows nothing about how the task gets solved.
- Harness , the program that actually runs the agent: a simple ReAct loop, a CLI agent like Codex or Kimi Code, or your own custom agent.
- Runtime , where execution happens: a local subprocess, Docker, or a remote sandbox like Prime Sandboxes or Modal.
Any taskset can run under any compatible harness, inside any runtime. This sounds simple, but it's a significant unlock: you can benchmark the same coding task against Codex, Mini-SWE-Agent, and your own agent without rewriting any environment code. You can also swap from a local Docker runtime during development to a horizontally-scaled remote sandbox for production runs, with no changes to the task definition.
The Interception Server: The Trick Behind Harness-Agnosticism
Making any harness work with any taskset requires solving a hard problem: agents like Claude Code speak Anthropic's API dialect, while Codex uses OpenAI Responses. The core abstraction that makes the v1 design work is the interception server -- a verifiers-managed HTTP server that proxies requests between the agent's runtime and the inference server.