Estimated reading time: 7 minutes · Last updated:
South Korea's Electronics and Telecommunications Research Institute (ETRI) has made TANGO, a no-code AI tool for building and deploying models, available as open source. ETRI announced the release on 9 November 2026. The tool lets people such as factory engineers, farm advisers and clinicians create AI models in a web browser, while TANGO handles training, optimization and deployment. Its latest version, TANGO 2, extends that workflow to generative AI and large language models, and ETRI has posted the LLMOps engine at its core on GitHub. The work comes from ETRI's Edge AI Platform Research Section, which has spent five years on the project. Partners in shipbuilding, agriculture and hospital radiology are already testing it, and ETRI is seeking an international standard for the approach.
Key takeaways
- Open-source release: ETRI said on 9 November 2026 that its no-code AI tool TANGO is now open source, with the LLMOps solution behind TANGO 2 published on GitHub.
- Five-year build: ETRI began TANGO in 2021. TANGO 1 handled computer vision such as image recognition, and TANGO 2 adds generative AI and large language models.
- Field partners: Acryl, Avnautics, Sure Data Lab, Seoul National University Hospital and Seoul National University Bundang Hospital are applying TANGO in commercial products, navigation, farming and radiology.
- First licensee: In April, the ICT Technology Commercialization Festival was the channel through which TANGO technology passed to Persona AI, and the company now builds on it for its GEN STATION agent platform.
- Standards push: ETRI wants ITU-T to give final approval to a no-code LLMOps standard in May 2027.
Table of contents
- Key takeaways
- How TANGO Folds Three Jobs Into One No-Code Tool
- What TANGO 2 Adds for Large Language Models
- Where TANGO Is Being Tested, From Ships to Chest Scans
- Can a Licensee and a Standard Carry TANGO Beyond ETRI?
- What could help or hold back TANGO
- What to be careful about
- Frequently asked questions
How TANGO Folds Three Jobs Into One No-Code Tool
Building an AI service has usually meant a relay between three groups of people. Domain experts on site assembled the data. AI specialists trained a model on it. Software developers then wired the result into working systems. Each handoff added time, and smaller firms often could not hire all three.
TANGO is ETRI's attempt to collapse that relay. A no-code tool is software that lets people build applications by choosing options in an interface rather than writing programs. TANGO applies that idea to the whole model lifecycle. Users pick the functions they want in a web browser. The tool then carries out training, optimization and deployment in the background.
The intended user is the person who understands the problem rather than the person who understands machine learning. That could be an inspector on a production line, a ship operator or a farm adviser. ETRI describes two savings: less time to get a model running, and less pressure to recruit scarce specialists. ETRI has not said how the tool helps non-specialists judge whether a model they trained is good enough. That question will decide how far TANGO spreads.
What TANGO 2 Adds for Large Language Models
ETRI started the project in 2021. The first version, TANGO 1, covered computer vision tasks such as image recognition. The institute has since widened it into TANGO 2, which also handles generative AI and large language models.
TANGO 2 is aimed at organisations that want to train an LLM on their own industry's data, manage that model over time and put it into a live service. Once a model is ready, the tool can optimize it and ship it to three kinds of target: the cloud, a company's in-house servers, or individual devices. That spread matters for a team whose focus is edge AI. Edge AI means models that run close to where data is produced rather than in a distant data centre.
The open-source release centres on LLMOps, which ETRI calls TANGO 2's core technology. LLMOps, short for large language model operations, is the tooling and practice used to develop, deploy and maintain LLMs once they are in use. ETRI's researchers have put this generative AI development and operations solution on GitHub. Outside developers can now inspect it, adapt it and build on it.
Where TANGO Is Being Tested, From Ships to Chest Scans
ETRI is working with partner institutions and companies to bring TANGO onto real sites. The most direct commercial link is with Acryl. The TANGO LLMOps tool is being tied into development of Jonathan, Acryl's commercial product, so that it can be sold straight away.
Two sector pilots run outside the lab. In shipbuilding and marine work, Avnautics is demonstrating autonomous navigation. Its AI reads data from multiple sensors to help vessels operate safely. In agriculture, Sure Data Lab is building a generative AI service for farms that draws on the experience and know-how of farming experts, and is working toward commercialization.
The medical work is the most data-heavy. Seoul National University Hospital has built technology that drafts radiology reports automatically. It learned from a large body of chest computed tomography (CT) images and the reports written about them. Together with Seoul National University Bundang Hospital, it intends to show a service for forecasting heart, lung and blood-vessel conditions. That is still a planned demonstration, not a tool in clinical use.
Can a Licensee and a Standard Carry TANGO Beyond ETRI?
Technology transfer has already begun. Persona AI is a company that specialises in agent AI. It took a transfer of TANGO technology in April at the ICT Technology Commercialization Festival, an event ETRI ran jointly with the Ministry of Science and ICT and the Institute of Information & Communications Technology Planning & Evaluation (IITP). Persona AI is building its business around an AI agent platform called GEN STATION. It has won a national AI award for its technology and is moving into physical AI, which pairs AI agents with robots.
ETRI is also pursuing an international standard for no-code LLMOps at the International Telecommunication Union's Telecommunication Standardization Sector (ITU-T). The institute has final approval targeted for May 2027. That is a goal ETRI has set, and no approval has been announced.
Cho Chang-sik, principal researcher in the Edge AI Platform Research Section, said ETRI would keep sharing its development results and field experience. He said the aim is an AI development environment that companies can use in practice. His colleagues on the demonstration work include Shin Ik-soo, Kim Hong-sook, Seok Jin-wook, Park Jae-bok and Lee Kyung-hee.
| Partner | Sector | Application | Status |
|---|---|---|---|
| Acryl | Commercial software | Linking TANGO LLMOps to its Jonathan product | Aimed at immediate commercialization |
| Avnautics | Shipbuilding and marine | AI autonomous navigation using sensor data | Demonstration |
| Sure Data Lab | Agriculture | Generative AI built on farming experts' knowledge | Pursuing commercialization |
| Seoul National University Hospital | Medical | Automatic radiology reports from chest CT | Technology developed |
| SNU Hospital and SNU Bundang Hospital | Medical | Cardiopulmonary and vascular disease prediction | Demonstration planned |
| Persona AI | Agent AI | GEN STATION platform, physical AI | Technology transferred in April |
What could help or hold back TANGO
The case for
- Publishing the LLMOps core on GitHub lets outside developers test and extend it without licensing talks with ETRI.
- Deployment to the cloud, in-house servers or individual devices covers firms that cannot send data off site.
- Live partner work in shipbuilding, agriculture and hospitals gives ETRI field cases to show at its seminar.
- ITU-T approval, if granted in May 2027, would give the no-code LLMOps approach an international reference point.
The case against
- ETRI has not described how non-specialists will validate the quality of models they train themselves.
- Most partner work is still at the demonstration or commercialization-pursuit stage rather than in routine use.
- The May 2027 ITU-T date is a target, and no approval has been announced.
What to be careful about
- Clinical prediction tools from Seoul National University Hospital and Seoul National University Bundang Hospital remain at the planned demonstration stage and are not cleared for patient care.
- Removing AI specialists from the workflow shifts responsibility for model errors onto domain staff, and ETRI has not said how TANGO supports model validation.
- Open-source projects from public institutes depend on continued maintenance, and ETRI has not set out a maintenance commitment for the GitHub release.
The bottom line
TANGO's open-source release matters less for any single feature than for who it hands the controls to. ETRI has spent five years turning a vision-only toolkit into a platform that can tune, manage and deploy large language models. It has now put the LLMOps core on GitHub for anyone to examine. The evidence so far is a set of partner pilots and one technology transfer, to Persona AI. There are not yet independent measurements of how well non-specialists fare with the tool. The next public tests are the 5th TANGO Open Seminar in Seoul on 11 November 2026 and the ITU-T standards decision ETRI is aiming for in May 2027.
What to watch
- ETRI holds the 5th TANGO Open Seminar on 11 November 2026 at the Korea Science and Technology Center in Seoul's Gangnam district, where it plans to show TANGO 2's main functions and field cases.
- Watch for ITU-T's decision on the no-code LLMOps standard, which ETRI is targeting for final approval in May 2027.
- Watch for Acryl to ship its Jonathan product with TANGO LLMOps built in; no launch date has been set.
Frequently asked questions
Is ETRI's TANGO free to use?
ETRI has released TANGO as open source and put the LLMOps solution at the core of TANGO 2 on GitHub. The specific licence terms have not been stated, so users should check the repository before commercial use.
Where can models built with TANGO 2 run?
TANGO 2 can optimize and deploy models to the cloud, to a company's in-house servers, or to individual devices. That range fits the edge AI focus of ETRI's Edge AI Platform Research Section.
Is TANGO being used in hospitals?
Seoul National University Hospital has developed automatic radiology report generation using chest CT images and their reports. It plans a disease-prediction demonstration with Seoul National University Bundang Hospital, which is not yet a clinical service.
Has any company licensed TANGO technology?
Persona AI received a TANGO technology transfer in April through the ICT Technology Commercialization Festival. It uses the technology around its GEN STATION AI agent platform and is moving into physical AI.
Related reading
