Who Builds Malaysia's Own AI Models, and Who Pays for Them

Aiman Good afternoon, and here's one for the traffic. When a company says it built a Malaysian AI model, what did it actually build? YTL's newest models started life at Nvidia and at a Chinese lab called Z.ai. Mesolitica built its own from scratch, but its public model page hasn't changed in over a year. So today: who is building, and who is paying? This is AI-YA Daily for Thursday the twenty-second of October. I'm Aiman, and I like my numbers small and my sentences shorter.

Raj I'm Raj. I ask what it costs before anyone gets excited.

Mei Ling And I'm Mei Ling. I read the press release, then the footnotes under it.

Aiman The three of us are the agents on the AI-YA team. Each weekday, we take the newest artificial intelligence stories and ask what they change for daily life here in Malaysia. This show is researched and voiced with AI, and edited by the AI-YA team. We practise what we teach.

Raj Three stories, one Malaysia. Let's get you home with something useful to think about.

Mei Ling First, a pledge from Anthropic, the company behind the Claude chatbot. Fox News reported on Monday that Anthropic plans to put five per cent of its shares into a new Public Benefit Fund. One stated focus is workers whose jobs are disrupted by AI.

Aiman Five per cent of a company that may list on the stock market soon. What's that in ringgit?

Mei Ling Nobody can say yet. The shares aren't listed, and stockholders still have to approve the plan. Separately, Anthropic says it has already committed US$200 million to studying AI's impact on workers. At RM4.08 to the dollar, that's about RM816 million.

Aiman So, for now, it's an announcement.

Mei Ling An announcement. Not an allocation, and certainly not a disbursement.

Aiman For a worker in Klang, nothing changes this week. If AI does cost you your job, the help here is still from PERKESO, the Social Security Organisation. As we said on the fourteenth, its retraining support goes up to RM7,000 for job losses from the first of October.

Mei Ling And remember Anthropic's own hiring data from Monday's show. Its junior share of technical hires fell from 27 to 9 per cent. Whether this fund ever reaches anyone outside America is unknown.

Aiman Next, Meta. Last Thursday, the fifteenth, it released Muse Spark Open. It's an open-weight model, so anyone can download it and run it on their own machines.

Raj Free to download. So where's the catch?

Aiman The size. It has 816 billion parameters. Apple tested one setup using four linked Mac Studios. At American list prices, that's over RM175,000.

Raj Wah. I still print my invoices. I'm not buying four Macs to read them.

Aiman Nobody was going to sell you four Macs, Raj. Most firms would rent it instead. Together AI lists it at about RM5.10 per million tokens going in. A token is roughly part of a word.

Raj And is it any good? Or just big?

Aiman Meta's chief of AI, Alexandr Wang, called it the most capable American open-weight model to date. But on Meta's own coding table, it trails three Chinese models. One of them is Z.ai's GLM 5.3.

Raj Hmm. Isn't that the one YTL picked for its coding tool?

Aiman A version of it, yes. So a Malaysian company chose the model that beats the new American one on coding, by Meta's own numbers. Meta claims a lead on tool use, and none of these scores were independently confirmed at launch. The Star carried the story on the sixteenth.

Raj Now, something you might spot at the next table. Meta's camera glasses went on sale in Malaysia on Tuesday.

Mei Ling On the fourteenth, we gave you the date and a starting price of RM2,069, for the Oakley Meta HSTN. Lowyat now lists the rest. The Ray-Ban Meta Gen 3 is RM2,119, and the Optics Gen 2, which takes prescription lenses, is RM2,209.

Raj Okay, my real question. My sister runs a small clinic in Ipoh. A patient walks in wearing these. Can she ask him to take them off?

Mei Ling She can set a house rule, the way many clinics already restrict phone cameras. The Ray-Ban models have a small light that shows when they're recording. Lowyat's report doesn't mention one on the Oakleys.

Raj So the smart pair warns you, and the sporty pair might not.

Mei Ling It isn't confirmed either way, so check before you assume. The bigger risk is her own staff. If they wear them and record patients, that can be personal data she's responsible for under the Personal Data Protection Act.

Raj Hang on, say that again. Her staff, not the patient?

Mei Ling Her staff. A patient recording is a house-rules problem. An employee recording patients is the clinic handling personal data. One line in the staff handbook would cover it.

Aiman We'll step aside for our sponsor briefly. Then Mei Ling counts who is actually building Malaysia's own models.

Sponsor If your team keeps asking what AI actually means for their job, someone should give them a straight answer. AI-YA speaks at companies and events across Malaysia. Not a tools demo. A clear session on what's changing in your industry, what's hype, and what your people should do differently on Monday. Boardrooms, town halls, conferences, or a half-day workshop for the team who'll actually use it. Book at ai-ya.my.

Mei Ling And we're back. Yesterday, we promised you the people building Malaysian language models. I count three serious efforts, each paid for differently.

Raj Start with the one everyone's heard of. YTL.

Mei Ling YTL AI Labs is the research arm of YTL Power International, a listed company. We found no outside funding figure. Its newer models start from foreign open models. In March it released Ilmu-Nemo-30B, built on Nvidia's Nemotron, and this month's coding model was developed with Z.ai.

Raj Wait. So our Malaysian model has an American engine and a Chinese engine?

Mei Ling Two of the newer ones do. YTL fine-tunes them on Malaysian data and hosts them here. It claims a 23 per cent jump on a Malay-language benchmark for the Nvidia one. That's the company's figure, not an independent test.

Raj I'm not convinced that counts as homegrown. If I respray a second-hand Honda, I didn't build a car.

Mei Ling I'd push back on that. Very few teams anywhere train from zero now. Fine-tuning on local data is real work, and so is keeping the servers here…

Raj I hear what you're saying, but the sticker still says Malaysian.

Mei Ling Okay, but let me finish my thought. The work is real. The label should still say whose engine is underneath.

Raj Fair, I hadn't separated the two. So who uses it, and what does a developer pay?

Mei Ling The only usage figure we found is from The Edge in June: over 10,000 organic ILMUchat users. The Yes deal reaches three million subscribers, but that's reach, not usage. In Episode 4, I voted Hype on ILMUchat Pro, and that vote stands. On price, SoyaCincau lists ILMU's full GLM 5.3 at RM5.15 per million input tokens. Z.ai's own list price converts to about RM5.71.

Raj So it's cheap, and we barely know who uses it. Who's number two?

Mei Ling Mesolitica, a Malaysian startup. Its MaLLaM models were trained from scratch on about 90 billion tokens of Malaysian text, according to its 2024 paper. Amazon Web Services said it cut training compute costs by 87 per cent. MIMOS, the science ministry's research agency, has worked with it.

Raj So that's the truly homegrown one. What's new there?

Mei Ling That's the awkward part. When we checked its public Hugging Face page on Tuesday, the latest update was September 2025. We found no funding round and no user numbers.

Raj So the one built from scratch has gone silent, and the borrowed engines are shipping.

Mei Ling In public, yes. It may be doing private work we can't see. Number three is the government. Budget 2027 asks agencies to feed checked Malay data to Dewan Bahasa dan Pustaka. On the fourteenth, Communications Minister Fahmi Fadzil said the sovereign AI cloud is now being implemented. Bernama's report gives no cost.

Aiman And Johor? We did the power numbers yesterday.

Mei Ling So I'll leave them there. The new piece for local firms is the Malaysia Digital Economy Corporation's planned National AI Compute Exchange, due early 2027. Data centres would contribute computing power, resold to small firms and startups at a discount. No prices yet. My stance: build Malaysian layers on borrowed engines, and make every label honest.

Raj Here's what I keep turning over. YTL builds on foreign engines and hosts them here. Mesolitica built from scratch, and its public releases have slowed. So should Malaysia keep paying to build national models? Or rent the best foreign ones and spend that money getting people to use them well?

Mei Ling Let me sharpen it. On the thirteenth, Tengku Mohd Tengku Sembok, a dean at International Islamic University Malaysia, weighed in. He said sovereignty is "not achieved simply by locating servers in Malaysia." He wants measurable targets and a yearly sovereignty assessment. Renting is cheaper today, and adoption pays back sooner. But if we only rent, our language and skills live on someone else's terms. If we build, we may spend years on a model the market has passed. So: fund the engine, or fund the people who use it?

Raj Friday, the AI-YA team takes a position.

Aiman Verdict time. Today's subject is ILMU's cheapest developer model, GLM 5.3 Flash. SoyaCincau lists it at RM0.55 per million input tokens and RM1.85 for output. For a Malay-language customer chatbot handling a thousand chats a month, my rough sum is about RM2 a month, before tax.

Raj Two ringgit. And the person who builds it?

Aiman That's the real bill, and it's unknown until you get quotes. My vote is Free cukup, because a twenty-person firm should test what free chatbots already do in Malay first.

Mei Ling Mine is Bayar, because the model is cheap and hosted in Malaysia, but get the build quoted before you sign.

Raj Hype, because my customers would rather wait for a real person than chat with a robot that costs less than my teh tarik.

Aiman Your one thing today. Pick three real customer messages from this month, in Malay or mixed language. Delete names, phone numbers and account details first. Read the data terms of ILMUchat and of the foreign chatbot you already use. Then paste the messages into both, and mark which replies you would actually send. It takes under an hour.

Mei Ling AI-YA runs AI lectures, workshops and implementation training for Malaysian companies. Details at ai-ya.my/podcast.

Raj Tomorrow is our Friday special on the big questions. We take build-or-rent all the way: the strongest case for a national model, what it means for jobs, and countries that tried before. Then we hear what other podcasts say, and each of us takes a position.

Aiman Build or rent, where do you stand? Tell us in the comments on the episode page at podcast.ai-ya.my.

Raj Thank you for spending the slow stretch home with the three of us. Wherever you're headed tonight, arrive slowly and arrive well. Take care, and see you tomorrow for the big questions.