Contact Sales
RLWRLD logo

RLWRLD

StartupSeoul, South KoreaFounded 2024

Develops RLDX-1, an open-weights vision-language-action model for dexterous, high-degree-of-freedom robot hands.

Overview

RLWRLD is a Seoul-based robotics AI startup founded in July 2024 by Jung-hee Ryu, founder of the computer-vision company Olaworks, which Intel acquired. In May 2026 it released RLDX-1, a vision-language-action model for dexterous manipulation, with weights on Hugging Face and code on GitHub. It has raised a $15M seed round (April 2025) and a $26M Seed 2 round (March 2026) from financial and strategic investors in Korea and Japan.

Sources: thebridge.jp (21 Apr 2025) · rlwrld.ai (19 Oct 2026) · pulse2.com (18 Apr 2025) · pulse2.com (3 Mar 2026)

Type
Startup
Headquarters
Seoul, South Koreatherobotreport.com, 11 May 2026
Founded
2024thebridge.jp, 21 Apr 2025
Website
rlwrld.ai

Key people

Funding history

Private company · $41M disclosed in USD across 2 of 2 listed rounds

Funding rounds for RLWRLD
DateRoundAmountLeadSource
18 Apr 2025Seed$15Mn/apulse2.com, 18 Apr 2025
3 Mar 2026Seed 2$26Mn/apulse2.com, 3 Mar 2026

Ranking

#20 of 104 model companies · score 68.5 / 100 · updated 19 Oct 2026

  • Published benchmarks 35%92%
  • Open weights 20%100%
  • Adoption 20%53%
  • Released models 15%20%
  • Disclosed funding 10%26%

Each bar shows an input from 0 to 100% before weighting. 1 of 1 listed models have open weights. Most-starred official repository: RLWRLD/RLDX-1 (355 stars, 19 Oct 2026).

How we rank model companies

The score (0–100) measures public, checkable evidence about a company's robot models. It does not measure private capability. A company that publishes no benchmark results and no weights can still have strong models. Every input is computed from the data on each company page, and every number there links to its source.

InputWeightHow it is measured
Published benchmarks35%For each scored benchmark, the company's best result divided by the best result any company has published, averaged over the benchmarks it reports. No published results scores 0. Results come only from the model's own paper, model card or the official benchmark leaderboard. Scored: LIBERO (4-suite average success, %); SimplerEnv Google Robot, visual matching (%); SimplerEnv Google Robot, variant aggregation (%); SimplerEnv WidowX / Bridge (%); CALVIN ABC→D (average length, 0–5); RoboArena (pairwise rating).
Open weights20%Share of the company's listed models whose weights the public can download.
Adoption20%GitHub stars on the company's most-starred official model repository, on a log scale relative to the most-starred repository in the list. Hugging Face downloads are shown on company pages but not scored, because gated repositories report zero.
Released models15%Number of publicly announced, named models, capped at 5.
Disclosed funding10%The sum of the funding rounds listed on the company's profile that were disclosed in USD, on a log scale from $10M to the largest sum in the list. Rounds reported only in other currencies are listed but not converted or counted. This input applies only to private startups and robot makers. For big-tech divisions, research labs and listed companies it is n/a, and its weight is shared across the other inputs in proportion.

Sources. arXiv papers and official model cards for benchmarks and parameters; official leaderboards (RoboArena); the GitHub and Hugging Face APIs for stars and downloads; company announcements and established press for funding. Missing values are shown as n/a, never estimated.

Updated 19 Oct 2026. Benchmark protocols differ between papers, for example in fine-tuning and number of trials, so small gaps between close scores are not meaningful.

Submit or correct data →

Open-weights models (1)

RLWRLD logo

RLDX-1

Open weightsVLA

Vision-language-action model for dexterous manipulation built on a Multi-Stream Action Transformer with vision, motion, memory and tactile/torque streams; VLM backbone Qwen3-VL 8B; mid-trained checkpoints are 8.1B and benchmark fine-tunes 6.9B parameters.

Released
7 May 2026
Parameters
8.1B
License
RLWRLD Model License v1.0 (weights); Apache-2.0 (code)
Embodiments
humanoid, single-arm, dexterous hand
GitHub stars
355 as of 19 Oct 2026
HF downloads (all time)
3.4k as of 19 Oct 2026
Published benchmark results for RLDX-1
BenchmarkResultSource
LIBEROAverage success rate across LIBERO-Spatial/Object/Goal/Long (%)97.8%huggingface.coModel card Performance table, RLDX-1-FT-LIBERO (fine-tuned on the 4-suite LIBERO set). As of 19 Oct 2026
SimplerEnv Google VMGoogle Robot visual matching average success (%)81.5%huggingface.coModel card Performance table, RLDX-1-FT-SIMPLER-GOOGLE. As of 19 Oct 2026
SimplerEnv Google VAGoogle Robot variant aggregation average success (%)77.4%huggingface.coModel card Performance table, RLDX-1-FT-SIMPLER-GOOGLE. As of 19 Oct 2026
SimplerEnv WidowXWidowX (Bridge) average success (%)71.9%huggingface.coModel card Performance table, RLDX-1-FT-SIMPLER-WIDOWX. As of 19 Oct 2026
RoboCasaRoboCasa Kitchen success rate (%)Shown, not scored70.6%huggingface.coModel card Performance table, RLDX-1-FT-ROBOCASA. As of 19 Oct 2026

Paper ↗Code ↗Hugging Face ↗Announcement ↗

Model companies ranking: #20, score 68.5 →VLA model guides →Benchmarks explained →Evaluate a policy on real hardware (RL² Eval) →Submit or correct data →

Robotics Center compiles this page from public sources. Listing a company does not imply a partnership or endorsement.