{"id":308561,"date":"2026-01-28T20:01:09","date_gmt":"2026-01-28T20:01:09","guid":{"rendered":"https:\/\/www.europesays.com\/ie\/308561\/"},"modified":"2026-01-28T20:01:09","modified_gmt":"2026-01-28T20:01:09","slug":"tiny-startup-arcee-ai-built-a-400b-parameter-open-source-llm-from-scratch-to-best-metas-llama","status":"publish","type":"post","link":"https:\/\/www.europesays.com\/ie\/308561\/","title":{"rendered":"Tiny startup Arcee AI built a 400B-parameter open source LLM from scratch to best Meta’s Llama"},"content":{"rendered":"

Many in the industry think the winners of the AI model market<\/a> have already been decided: Big Tech will own it (Google, Meta, Microsoft, a bit of Amazon) along with their model makers of choice, largely OpenAI and Anthropic.\u00a0<\/p>\n

But tiny 30-person startup Arcee AI<\/a> disagrees. The company just released a truly and permanently open (Apache license) general-purpose, foundation model called Trinity, and Arcee claims that at 400B parameters, it is among the largest open source foundation models ever trained and released by a U.S. company.<\/p>\n

Arcee says Trinity compares to Meta\u2019s Llama 4 Maverick 400B, and Z.ai\u2019s GLM-4.5, a high-performing open source model from China\u2019s Tsinghua University, according to benchmark tests conducted using base models (very little post-training).<\/p>\n

\"ArceeArcee AI benchmarks for its Trinity large LLM (preview version, base model)Image Credits:<\/strong>Arcee AI<\/p>\n

Like other state-of-the-art (SOTA) models, Trinity is geared for coding and multi-step processes like agents. Still, despite its size, it\u2019s not a true SOTA competitor yet because it currently supports only text. <\/p>\n

More modes are in the works \u2014 a vision model is currently in development, and a speech-to-text version is on the roadmap, CTO Lucas Atkins told TechCrunch (pictured above, on the left). In comparison, Meta\u2019s Llama 4 Maverick is already multi-modal, supporting text and images.<\/p>\n

But before adding more AI modes to its roster, Arcee says, it wanted a base LLM that would impress its main target customers: developers and academics. The team particularly wants to woo U.S. companies of all sizes away from choosing open models from China.\u00a0<\/p>\n

\u201cUltimately, the winners of this game, and the only way to really win over the usage, is to have the best open-weight model,\u201d Atkins said. \u201cTo win the hearts and minds of developers, you have to give them the best.\u201d<\/p>\n

Techcrunch event<\/p>\n

\n\t\t\t\t\t\t\t\t\tSan Francisco
\n\t\t\t\t\t\t\t\t\t\t\t\t\t|
\n\t\t\t\t\t\t\t\t\t\t\t\t\tOctober 13-15, 2026\n\t\t\t\t\t\t\t<\/p>\n

The benchmarks show that the Trinity base model, currently in preview while more post-training takes place, is largely holding its own and, in some cases, slightly besting Llama on tests of coding and math, common sense, knowledge, and reasoning.<\/p>\n

The progress Arcee has made so far to become a competitive AI Lab is impressive. The large Trinity model follows two previous small models<\/a> released in December: the 26B-parameter Trinity Mini, a fully post-trained reasoning model for tasks ranging from web apps to agents, and the 6B-parameter Trinity Nano, an experimental model designed to push the boundaries of models that are tiny yet chatty.\u00a0\u00a0<\/p>\n

The kicker is, Arcee trained them all in six months for $20 million total, using 2,048 Nvidia Blackwell B300 GPUs. This out of the roughly $50 million the company has raised so far, said founder and CEO Mark McQuade (pictured above, on the right).\u00a0<\/p>\n

That kind of cash was \u201ca lot for us,\u201d said Atkins, who led the model-building effort. Still, he acknowledged that it pales in comparison to how much bigger labs are spending right now.<\/p>\n

The six-month timeline \u201cwas very calculated,\u201d said Atkins, whose career before LLMs involved building voice agents for cars. \u201cWe are a younger startup that\u2019s extremely hungry. We have a tremendous amount of talent and bright young researchers who, when given the opportunity to spend this amount of money and train a model of this size, we trusted that they\u2019d rise to the occasion. And they certainly did, with many sleepless nights, many long hours.\u201d\u00a0<\/p>\n

McQuade, previously an early employee at open source model marketplace Hugging Face, says Arcee didn\u2019t start out wanting to become a new U.S. AI lab:\u00a0The company was originally doing model customization for large enterprise clients like SK Telecom.\u00a0<\/p>\n

\u201cWe were only doing post-training. So we would take the great work of others: We would take a Llama model, we would take a Mistral model, we would take a Qwen model that was open source, and we would post-train it to make it better\u201d for a company\u2019s intended use, he said, including doing the reinforcement learning.\u00a0<\/p>\n

But as their client list grew, Atkins said, the need for their own model was becoming a necessity, and McQuade was worried about relying on other companies. At the same time, many of the best open models were coming from China, which U.S. enterprises were leery of, or were barred from using.\u00a0<\/p>\n

It was a nerve-wracking decision. \u201cI think there\u2019s less than 20 companies in the world that have ever pre-trained and released their own model\u201d at the size and level that Arcee was gunning for, McQuade said.\u00a0<\/p>\n

The company started small at first, trying its hand at a tiny, 4.5B model created in partnership with training company DatologyAI. The project\u2019s success then encouraged bigger endeavors.\u00a0<\/p>\n

But if the U.S. already has Llama, why does it need another open weight model? Atkins says by choosing the open source Apache license, the startup is committed to always keeping its models open. This comes after Meta CEO Mark Zuckerberg last year indicated his company might not always<\/a> make all of its most advanced models open source.\u00a0<\/p>\n

\u201cLlama can be looked at as not truly open source as it uses a Meta-controlled license with commercial and usage caveats,\u201d he says. This has caused some open source organizations to claim<\/a> that Llama isn\u2019t open source compliant at all.<\/p>\n

\u201cArcee exists because the U.S. needs a permanently open, Apache-licensed, frontier-grade alternative that can actually compete at today\u2019s frontier,\u201d McQuade said.<\/p>\n

All Trinity models, large and small, can be downloaded for free. The largest version will be released in three flavors. Trinity Large Preview is a lightly post-trained instruct model, meaning it\u2019s been trained to follow human instructions, not just predict the next word, which gears it for general chat usage. Trinity Large Base\u00a0is the base model without post-training.<\/p>\n

Then we have TrueBase,\u00a0a model with any instruct data or post training so enterprises or researchers that want to customize it won\u2019t have to unroll any data, rules, or assumptions.<\/p>\n

Arcee AI will eventually offer a hosted version of its general-release model for, it says, competitive API pricing. That release is up to six weeks away as the startup continues to improve the model\u2019s reasoning training.<\/p>\n

API pricing for Trinity Mini is $0.045 \/ $0.15, and there is a rate-limited\u00a0free tier available, too.\u00a0Meanwhile, the company still sells post-training and customization options.\u00a0<\/p>\n","protected":false},"excerpt":{"rendered":"Many in the industry think the winners of the AI model market have already been decided: Big Tech…\n","protected":false},"author":2,"featured_media":308562,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[261],"tags":[291,151162,289,290,18,45280,19,17,151163,8211,82,41729],"class_list":{"0":"post-308561","1":"post","2":"type-post","3":"status-publish","4":"format-standard","5":"has-post-thumbnail","7":"category-artificial-intelligence","8":"tag-ai","9":"tag-arcee-ai","10":"tag-artificial-intelligence","11":"tag-artificialintelligence","12":"tag-eire","13":"tag-foundation-models","14":"tag-ie","15":"tag-ireland","16":"tag-llama-4","17":"tag-open-source-ai","18":"tag-technology","19":"tag-trinity"},"share_on_mastodon":{"url":"https:\/\/pubeurope.com\/@ie\/115974509886356995","error":""},"_links":{"self":[{"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/posts\/308561","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/comments?post=308561"}],"version-history":[{"count":0,"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/posts\/308561\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/media\/308562"}],"wp:attachment":[{"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/media?parent=308561"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/categories?post=308561"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.europesays.com\/ie\/wp-json\/wp\/v2\/tags?post=308561"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}