Skip to content
@Valen-Team

Valen-Team

Valen logo

Valen · 万澜

Fast Decisions. Continuous Evolution.
Fast decision models · Recursive self-improvement (RSI) · Agents

English · 简体中文

Website · Playground · Documentation · News · Hugging Face · Leaderboard

The Valen ecosystem: Valen-Decision-1.5 takes text, images and video with questions and returns typed answers (predicate, choice, score, bbox). Around it are the SDKs, agent skill and workflow integrations, the runtimes from the hosted API to vLLM, SGLang, ONNX Runtime, llama.cpp and Ollama, evaluation with JevalKit, and Roller and Awesome-RSI for agents and recursive self-improvement.


Valen is an open-source organization bringing together fast decision models, recursive self-improvement (RSI), and agents. We develop and bring together models, tools, research resources, and applications to help agents make decisions quickly, act in their environments, and improve through experience and feedback.

Our goal is to connect efficient decision-making with agents that can learn and adapt over time. We welcome researchers and developers working across these areas to build, evaluate, and share their work here.

What We Work On

  • Fast decision models — low-latency decisions from text, images, and video, with structured outputs that agents can use to select actions, route tasks, and assess states.
  • Recursive self-improvement (RSI) — methods for learning from experience, reusing memory and skills, improving tools and agent workflows, and updating models. We care about reproducible evidence that these changes lead to better performance over time.
  • Agents — agent frameworks, tools, and applications that connect perception, planning, decision-making, and execution in interactive environments.

These directions inform one another: decision models help agents choose actions; agent runs produce experience and feedback; RSI methods explore how to turn that feedback into improvements. Evaluation helps establish whether those improvements hold up.

Open-Source Projects

Project What it does Explore
Valen-Decision A multimodal System One decision model: text, images, and video in; structured decision probabilities out. Valen-Decision-1.5 comes in five sizes from 0.8B to 27B and answers bbox questions. Includes training, fine-tuning, inference, and evaluation. Try it · Playground · Docs & API
SDKs & integrations Call Valen-Decision from your own stack: the Python SDK, the JavaScript SDK, the C++ text integration, an agent skill for Claude Code, Cursor and Codex, and workflow integrations for n8n, LangChain, Dify and Coze. ONNX & SDK guide · Cookbook
JevalKit An evaluation toolkit for Jev-like decision models, with unified model interfaces, prepared datasets, benchmark scoring, and a public leaderboard. Docs · Leaderboard
Roller Agent trajectory collection across models and Harnesses. Runs tool-using agents such as Claude Code, Codex and OpenClaw in disposable Docker containers, grades benchmark tasks in the container where they ran, and stores the native session, a unified trace and SFT training data. Covers seven Harnesses and five benchmarks. Quick start · Docs · PyPI
Awesome-RSI A curated collection of papers, code, benchmarks, datasets, and models for recursive self-improvement, organized around experience accumulation, system modification, and model parameter updates. Papers · Projects

Our public projects cover decision models, evaluation, agent trajectory collection, and RSI research resources. We plan to expand the organization with more agent and RSI projects, sharing implementations, experiments, and reusable tools as they become available.

Models, Datasets & Evaluation

New: Valen-Decision-1.5 is released in five sizes, as native weights, ONNX and GGUF. Get it from the Valen-Decision-1.5 collection.

Browse all releases on Hugging Face, or start with the documentation to use Valen-Decision and run evaluations with JevalKit.

The Name

Valen is named after evolutionary biologist Leigh Van Valen, who proposed the Red Queen hypothesis: organisms must continually adapt as the species around them evolve. Its literary inspiration comes from the Red Queen's race in Through the Looking-Glass.

For us, the name connects speed with continuous evolution: making decisions quickly while continuing to learn and adapt.

Our Chinese name, 万澜 (Wàn Lán), evokes countless waves in motion. The wave-shaped V in our logo carries that idea forward.

Build With Us

We welcome contributions and collaboration across decision models, RSI, and agents:

  • Decision models and evaluation — improve Valen-Decision, add models or benchmarks to JevalKit, or share data and reproducible results.
  • RSI research and methods — contribute approaches to learning from experience, evolving agent systems, or updating models; suggest papers and implementations for Awesome-RSI.
  • Agents and applications — add Harnesses or benchmarks to Roller, contribute agent frameworks, tools, integrations, and applications, or propose new agent and RSI projects for the organization.

Open an issue or submit a pull request in the relevant repository. Documentation, examples, and translations are welcome too.

Popular repositories Loading

  1. Valen Valen Public

    Open multimodal decision models, 0.8B to 27B, for text, images and video. Decide with probabilities over your own candidates and locate with bounding boxes, in place of generated text. Free API, ON…

    Python 916 88

  2. Roller Roller Public

    Turn agent runs into training data. Roller rolls out Claude Code, Codex, OpenClaw and four other Harnesses in disposable Docker containers, grades benchmark tasks in place, and exports SFT-ready tr…

    Python 57

  3. Awesome-RSI Awesome-RSI Public

    A curated list of papers, code, benchmarks and models on recursive self-improvement: AI systems that learn from experience, rewrite their own agent software, or update their weights.

    30 5

  4. JevalKit JevalKit Public

    Evaluate typed decision models such as Jev and Valen on text, image and video benchmarks, with one runner and a public leaderboard.

    Python 11 6

  5. .github .github Public

    Valen-Team

    Python 1

  6. valen-team.github.io valen-team.github.io Public

    Source for the Valen Team website and documentation.

    MDX 1

Repositories

Showing 10 of 12 repositories
  • .github Public

    Valen-Team

    Valen-Team/.github's past year of commit activity
    Python 1 0 0 0 Updated Nov 16, 2026
  • Roller Public

    Turn agent runs into training data. Roller rolls out Claude Code, Codex, OpenClaw and four other Harnesses in disposable Docker containers, grades benchmark tasks in place, and exports SFT-ready trajectories that keep reasoning, images and compacted-context segments.

    Valen-Team/Roller's past year of commit activity
    Python 57 Apache-2.0 0 0 0 Updated Nov 16, 2026
  • valen-team.github.io Public

    Source for the Valen Team website and documentation.

    Valen-Team/valen-team.github.io's past year of commit activity
    MDX 1 Apache-2.0 0 0 0 Updated Nov 16, 2026
  • Valen Public

    Open multimodal decision models, 0.8B to 27B, for text, images and video. Decide with probabilities over your own candidates and locate with bounding boxes, in place of generated text. Free API, ONNX, GGUF, vLLM and SGLang; fine-tuning code included.

    Valen-Team/Valen's past year of commit activity
    Python 916 Apache-2.0 88 0 0 Updated Nov 16, 2026
  • JevalKit Public

    Evaluate typed decision models such as Jev and Valen on text, image and video benchmarks, with one runner and a public leaderboard.

    Valen-Team/JevalKit's past year of commit activity
    Python 11 Apache-2.0 6 0 0 Updated Nov 15, 2026
  • valen-sdk-cpp Public

    C++ integration for Valen ONNX inference.

    Valen-Team/valen-sdk-cpp's past year of commit activity
    C++ 0 Apache-2.0 0 0 0 Updated Nov 15, 2026
  • valen-integrations Public

    Valen for n8n, Dify, Coze and LangChain: typed questions and probability-based branching in workflow and agent platforms

    Valen-Team/valen-integrations's past year of commit activity
    Python 1 MIT 0 0 0 Updated Nov 15, 2026
  • skills Public

    Agent skills for building with Valen, the multimodal System One decision model (Claude Code and Codex plugins, evals)

    Valen-Team/skills's past year of commit activity
    Python 1 Apache-2.0 0 0 0 Updated Nov 15, 2026
  • Awesome-RSI Public

    A curated list of papers, code, benchmarks and models on recursive self-improvement: AI systems that learn from experience, rewrite their own agent software, or update their weights.

    Valen-Team/Awesome-RSI's past year of commit activity
    30 5 0 0 Updated Nov 14, 2026
  • valen-sdk-js Public

    JavaScript SDK for Valen in Node.js and browsers.

    Valen-Team/valen-sdk-js's past year of commit activity
    JavaScript 0 Apache-2.0 0 0 0 Updated Nov 11, 2026

People

This organization has no public members. You must be a member to see who’s a part of this organization.