ClaimsMark Zuckerberg

“We're training three versions: an 8 billion parameter model and a 70 billion, which we're releasing today, and a 405 billion dense model, which is still training.”

Mark Zuckerberg, Meta CEO 18 April 2024 Mark Zuckerberg - Llama 3, Open Sourcing $10b Models, & Caesar Augustus (Dwarkesh Podcast)

Factual AI capabilities

Meta is training three Llama 3 versions: 8B and 70B parameter models, released that day, and a 405B dense model still in training.

Unresolvable The public record could not settle it.

Factual: Statements about the present or the past, checked against the record: did they hold up?

The second judge would not resolve this claim; the verdict shown is the first judge's.

Said for an organisation they lead or work for; it counts on both records.

What would count as true or false

Judge as of 2024-04-18. True: Meta's official Llama 3 announcement or model cards show that 8B and 70B parameter models were released on or about 2024-04-18, and that a model of about 400B parameters (dense) was still in training at that time. Mostly true: everything holds except a minor detail of a number or description (for example, Meta described the larger model as 'over 400B' rather than exactly 405B). Mixed: the release of the two models is accurate, but the claim about the third model is materially wrong. Mostly false: only part of it holds (for example, the models released were of different sizes, or they were released on a different date). False: no Llama 3 models were released that day.

Set before the research began.

Evidence (4)

  1. E1 Introducing Meta Llama 3: The most capable openly available LLM to date Meta AI Own record

    “Our largest models are over 400B parameters and, while these models are still training, our team is excited about how they’re trending.”
  2. E2 llama3/MODEL_CARD.md at main · meta-llama/llama3 GitHub (meta-llama/llama3 repository) · 18 April 2024 Own record

    “Model Release Date April 18, 2024.”
  3. E3 Welcome Llama 3 - Meta's new open LLM Hugging Face · 18 April 2024

    “Meta’s Llama 3, the next iteration of the open-access Llama family, is now released and available at Hugging Face.”
  4. E4 The good and the bad after 3 days of Meta's Llama 3 | Product Hunt Product Hunt (story by Kyle Corbitt, founder of OpenPipe) · 24 April 2024

    “The 405B model is still in training but weights for both of the smaller sizes are available.”

Reasoning

The evidence shows that Llama 3 was released on April 18, 2024 [E2][E3] and that a model of over 400B parameters was still training [E1][E4]. No evidence item states that the released models were 8B and 70B. E4 refers only to "both of the smaller sizes". Under the criteria, released models of different sizes would make the claim mostly false, so the evidence does not settle the verdict.

Cite this claim

Mark Zuckerberg, “Meta is training three Llama 3 versions: 8B and 70B parameter models, released that day, and a 405B dense model still in training.” (said 18 April 2024). Verdict: unresolvable. Reliability API, claim c_08235742d102, method 0.1.0+3f84d14d, dataset release v1.1.2. https://epistacklabs.com/claims/c_08235742d102

This claim is in dataset release v1.1.2, with the same verdict.

Reply to this claim

For Mark Zuckerberg, or someone speaking for them: a correction or counter-evidence. Replies are read before anything is shown; an approved reply shows your name, role, date, statement and link. Your email address is never shown and is deleted 30 days after the decision (privacy).

More factual claims by Mark Zuckerberg

  1. Resolved true Factual 15 September 2026
    “It is bigger than avocado.”

    Meta's Watermelon model is bigger than its Avocado model.

  2. Mostly true Factual 30 July 2025
    “We're seeing early progress with the launch of our AI video editing tools across Meta AI and our new Edits app, and there's a lot more to do here.”

    Meta launched AI video editing tools across Meta AI and its new Edits app.

  3. Mostly true Factual 23 July 2024
    “This year, Llama 3 is competitive with the most advanced models and leading in some areas.”

    In 2024, Llama 3 is competitive with the most advanced models and leading in some areas.

  4. Mostly true Factual 18 April 2024
    “One that people are going to find pretty wild is that it now generates high quality images so quickly that it actually generates it as you're typing and updates it in real time.”

    Meta AI now generates high-quality images fast enough to generate and update them in real time as the user types.

  5. Resolved true Factual 18 April 2024
    “we didn't prioritize coding because people aren't going to ask Meta AI a lot of coding questions in WhatsApp.”

    Meta did not prioritize coding when working on Llama 2.

All of them

Verdicts are made automatically; how verdicts are made.