cd /news/artificial-intelligence/south-korea-cuts-startup-motif-from-… · home › topics › artificial-intelligence › article
[ARTICLE · art-100798] src=startupfortune.com ↗ pub= topic=artificial-intelligence verified=true sentiment=· neutral

South Korea Cuts Startup Motif From Its AI Squid Game Contest

South Korea's Ministry of Science and ICT cut Motif Technologies from its Sovereign AI Foundation Model project after the first-phase evaluation, advancing LG AI Research Institute, SK Telecom, and Upstage, while opening a supplementary call for one more team. Motif's consortium was later selected on February 20, 2026, to rejoin the race, despite its Motif 3 Beta model scoring 44 on Artificial Analysis' Intelligence Index, matching DeepSeek's V4 Pro. The ministry's evaluation framework allocated 40 points to benchmarks, 35 to expert review, and 25 to user testing, with LG AI Research leading all categories.

read5 min views25 publishedAug 18, 2026
South Korea Cuts Startup Motif From Its AI Squid Game Contest
Image: Startupfortune (auto-discovered)

South Korea's state-backed race to build a homegrown AI model has become less about one leaderboard score and more about whether a team can prove it can ship a national system.

Motif Technologies is the uncomfortable contestant in Korea's AI contest because it has the thing every startup wants: a model that can stand on a public benchmark beside far larger rivals. That doesn't mean it wins. In a government program built to choose two national AI teams, benchmark numbers are only part of the fight.

The Ministry of Science and ICT launched the Sovereign AI Foundation Model project in 2025 to reduce Korea's dependence on foreign foundation models and push the country toward its AI G3 goal. The format is deliberately harsh. Five teams start, then the field is cut through staged evaluations until two winners remain. Bloomberg gave it the nickname 'AI Squid Game' after Netflix's survival drama, and frankly, the name works because the prize is real: government funding, GPU access, data support and the political blessing that comes with being treated as part of national AI infrastructure.

Motif wasn't in the surviving group after the first major cut. MSIT said LG AI Research Institute, SK Telecom and Upstage advanced after the first-phase evaluation, while the ministry opened a supplementary call for one more team. Korea Times reported that Motif's consortium was selected on February 20, joining the three teams already in the race. That is how a young company ended up standing beside LG and SK Telecom in a contest built around sovereign AI.

The Scorecard Is Not A Leaderboard #

If you only look at Artificial Analysis, Motif has a clean argument. Artificial Analysis lists Motif 3 Beta as a 314 billion parameter reasoning model released on July 14, 2026, with a score of 44 on its Intelligence Index. Yonhap also reported on July 21 that Motif said the model scored 44, matching DeepSeek's latest V4 Pro on that index.

[Relay shuts down as founder Jacob Bank returns to Google to lead Chrome AI](https://startupfortune.com/relay-shuts-down-as-founder-jacob-bank-returns-to-google-to-lead-chrome-ai/)

Relay, the AI workflow automation startup built to be an AI-era Zapier, is shutting down after free accounts lost access on August 15 and paid accounts will follow on September 14. Founder Jacob Bank is rejoining Google as VP of Product for Chrome, his second return to the company after it bought his earlier startup Timeful in 2015. - AI automation startup Relay shuts down permanently - founder Jacob Bank returns to Google Chrome

That is not a small result.

But you shouldn't confuse a third-party leaderboard with a government procurement scorecard. MSIT's first-phase framework split 100 points across three categories: 40 for benchmark evaluation, 35 for expert evaluation and 25 for user evaluation. The ministry said the user portion involved a panel of 49 expert users, including AI startup CEOs, who tested websites using the models for real-world usefulness and reasoning cost efficiency. LG AI Research led every major first-phase category, with 33.6 out of 40 on benchmarks, 31.6 out of 35 from experts and a perfect 25 in user evaluation.

That tells you what Korea is really buying. It wants model performance, yes, but it also wants a team that can explain its training process, defend its sovereignty claims, connect the model to working services and keep improving under pressure. A startup can beat a bigger company on one public index and still face a harder question from the state: can you run this at national scale?

Why Motif Still Matters #

Motif's February return was not a courtesy invite. The official Korean policy briefing said the company was chosen for its experience designing AI models with independent architecture and for reaching competitive performance despite limited parameters and data. The consortium included Moreh, KAIST, Seoul National University, Mathpresso and other partners, according to Korean government and local media reports. That matters because sovereign AI is not only a model file. It is training infrastructure, data work, application partners and enough operational depth to survive the next cut.

Look at the rest of the field. SK Telecom can bring telecom data, services and an existing consumer AI channel. LG AI Research has K-EXAONE and the backing of one of Korea's biggest industrial groups. Upstage has Solar, a strong local AI brand and a record of moving fast in open-weight models. Motif has technical credibility and a partner network. That is useful. It is also a thinner shield than a conglomerate balance sheet.

According to Aju Press, the second-phase evaluation was approaching in early August, with SK Telecom and Upstage showing strong scores across several common benchmarks after the participating companies released models on Hugging Face in July. The same report said LG's K-EXAONE 2.0 remained the largest model among the group at 750 billion parameters, while the public benchmark picture varied by category. You can see the shape of the contest there. No single number settles it.

The winners are expected to be picked by the end of the year, Korea Times reported, and those two teams will continue receiving state support to push Korea's sovereign AI effort forward. For Motif, the useful lesson is blunt: proving technical quality gets you into the room, but it doesn't own the room. Korea is testing whether a model can become infrastructure. That is a much harder test than a leaderboard.

Groq Raises Another $350 Million to Build Out Its Nvidia-Powered Neocloud Groq raised $350 million on August 17 at a $3.5 billion valuation, the second big check in two months as it completes its pivot from selling AI chips to renting out Nvidia GPU capacity. The move follows Nvidia's $20 billion licensing deal for Groq's LPU technology and the departure of founder Jonathan Ross. - groq raises 350 million funding round valuation - groq neocloud nvidia gpu computing rental service

Also read: Groq Raises $350 Million to Trade Its Chip Ambitions for a Cloud Business • Relay shuts down as founder Jacob Bank returns to Google to lead Chrome AI • Anthropic's Annualized Revenue Rockets to $65 Billion Ahead of Its IPO

── more in #artificial-intelligence 4 stories · sorted by recency
── more on @ministry of science and ict 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
→ Live at https://your-agent.zahid.host ✓
Get free account → Pricing
from €0/mo · no card required
LIVE [news/south-korea-cuts-sta…] indexed:0 read:5min 2026-08-18 · —