ClaimsPeople

Mark Zuckerberg

Mark Zuckerberg is a technology entrepreneur who co-founded Facebook and serves as founder, chairman and chief executive officer of Meta Platforms.

1,575 claims collected, 292 researched, 90 resolved. Report of 5 November 2026, method 0.1.0+3f84d14d.

Scores by kind of claim

Each kind of claim is scored on its own, out of 100: the share that held up, with the range it likely falls in. Fewer resolved claims give a wider range.

  • Factual claims

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

    80 74 86

    High confidence, 70 factual claims resolved. From 15 documents, 2017 to 2026; 15 documents sampled from 31 candidates found.

    Not scored: 40 unresolvable, 41 checkable only against their own records.

  • Predictions

    Statements about events outside the person's control, checked once they fell due: did they come true?

    Insufficient record: 4 predictions resolved, 5 needed for a score.

    Not scored: 98 unresolvable. 12 more fall due later.

    Does not yet allow for how likely each prediction was when made.

  • Promises

    Commitments to the person's or organisation's own future action, with something to deliver: was it delivered?

    85 74 94

    Medium confidence, 16 promises resolved. From 9 documents, 2019 to 2026; 15 documents sampled from 31 candidates found.

    Not scored: 6 unresolvable, 3 not yet due.

  • Guidance

    A company's own figures for its results over a period, checked against what it then reported.

    Insufficient record: 0 guidance claims resolved, 5 needed for a score.

    Not scored: 2 not yet due.

All types combined: 81 (likely 76 to 86, high confidence), from 86 claims: 70 factual claims, 16 promises. It mixes kinds of claim in different proportions for different people, so compare people kind by kind.

Mark Zuckerberg's claims

374 claims, sorted by newest said. Choose a claim to read it in full.
VerdictKindSaidClaimTopic
Mostly true Factual 23 Jul 2024 Meta is building internal teams to enable developers and partners to use Llama and is actively building partnerships with ecosystem companies. Factual · 23 Jul 2024 Their own work or organisation
Unresolvable Factual 23 Jul 2024 As of July 2024, several tech companies are developing leading closed AI models. Factual · 23 Jul 2024 AI capabilities
Resolved true Factual 23 Jul 2024 Zuckerberg said that one of his formative experiences has been building Meta's services constrained by what Apple allows on its platforms. Factual · 23 Jul 2024 Economics and markets
Not researched Factual 23 Jul 2024 Because the Llama models are open, anyone can test them for themselves. Factual · 23 Jul 2024 AI safety and alignment
Mixed Factual 23 Jul 2024 Developers can run inference on Llama 3.1 405B on their own infrastructure at roughly 50% the cost of using closed models like GPT-4o, for both user-facing and offline inference tasks. Factual · 23 Jul 2024 Economics and markets
Own records only Factual 23 Jul 2024 Meta has saved billions of dollars by releasing its server, network and data center designs via the Open Compute Project and having supply chains standardize on them. Factual · 23 Jul 2024 Economics and markets
Not researched Factual 23 Jul 2024 In 2023, Llama 2 was only comparable to an older generation of models behind the frontier. Factual · 23 Jul 2024 AI capabilities
Own records only Factual 23 Jul 2024 Meta manages security on its social networks using more robust AI systems that identify and stop threats from less sophisticated actors, who often use smaller-scale AI systems. Factual · 23 Jul 2024 AI safety and alignment
Mixed Factual 23 Jul 2024 Llama is already leading on openness, modifiability, and cost efficiency. Factual · 23 Jul 2024 Their own work or organisation
Not researched Factual 23 Jul 2024 On 2024-07-23 Meta released Llama 3.1 405B, described as the first frontier-level open source AI model, plus new Llama 3.1 70B and 8B models. Factual · 23 Jul 2024 Their own work or organisation
Not researched Factual 23 Jul 2024 Meta's safety process includes rigorous testing and red-teaming to assess whether its models can cause meaningful harm, aiming to mitigate risks before release. Factual · 23 Jul 2024 AI safety and alignment
Mostly true Factual 23 Jul 2024 In 2024, Llama 3 is competitive with the most advanced models and leading in some areas. Factual · 23 Jul 2024 AI capabilities
Not researched Factual 23 Jul 2024 With past Llama models, Meta developed them for itself and then released them without focusing much on building a broader ecosystem. Factual · 23 Jul 2024 Their own work or organisation
Not researched Factual 23 Jul 2024 Selling access to AI models is not Meta's business model. Factual · 23 Jul 2024 Economics and markets
Not researched Factual 23 Jul 2024 Amazon, Databricks, and NVIDIA are launching full suites of services supporting developers fine-tuning and distilling their own models. Factual · 23 Jul 2024 Economics and markets
Unresolvable Factual 23 Jul 2024 It is well-accepted that open source software tends to be more secure because it is developed more transparently. Factual · 23 Jul 2024 Technology
Unresolvable Factual 23 Jul 2024 Linux is the industry standard foundation for cloud computing and for the operating systems running most mobile devices. Factual · 23 Jul 2024 Technology
Not researched Factual 23 Jul 2024 Open source AI models are quickly closing the performance gap with leading closed models. Factual · 23 Jul 2024 AI capabilities
Not researched Factual 18 Apr 2024 Meta was infrastructure-constrained in catching up to TikTok as quickly as it wanted. Factual · 18 Apr 2024 AI compute, hardware and data
Not researched Factual 18 Apr 2024 Meta's Llama license is largely permissive open source but limits use by the largest companies. Factual · 18 Apr 2024 AI policy and governance
Not researched Factual 18 Apr 2024 Meta expected the Llama 3 model to plateau more during training on about 15 trillion tokens, but it was still learning at the end. Factual · 18 Apr 2024 AI capabilities
Unresolvable Factual 18 Apr 2024 Meta is training three Llama 3 versions: 8B and 70B parameter models, released that day, and a 405B dense model still in training. Factual · 18 Apr 2024 AI capabilities
Not researched Factual 18 Apr 2024 Meta did not conceive of FAIR as a product but as a research group. Factual · 18 Apr 2024 Their own work or organisation
Own records only Factual 18 Apr 2024 With unconnected content recommendations, Meta's pool of candidate content expanded from thousands to hundreds of millions. Factual · 18 Apr 2024 Technology
Unresolvable Factual 18 Apr 2024 Zuckerberg said he thinks Meta had the extra GPU capacity because it was working on Reels. Factual · 18 Apr 2024 Their own work or organisation
Not researched Factual 18 Apr 2024 Llama 3 8B and 70B models lead their scale (size class) on benchmarks. Factual · 18 Apr 2024 AI capabilities
Own records only Factual 18 Apr 2024 Meta concluded it must solve general intelligence and increased its investment to do so. Factual · 18 Apr 2024 Their own work or organisation
Not researched Factual 18 Apr 2024 Llama 3 8B is nearly as powerful as the largest Llama 2 model. Factual · 18 Apr 2024 AI capabilities
Own records only Factual 18 Apr 2024 Zuckerberg decided Meta should order enough GPUs for Reels and ranking needs, plus double that amount. Factual · 18 Apr 2024 Their own work or organisation
Not researched Factual 18 Apr 2024 Meta AI is integrating Google and Bing for real-time knowledge. Factual · 18 Apr 2024 AI capabilities
Not researched Factual 18 Apr 2024 Over the previous 18 months it emerged that training models on coding helps them across many non-coding domains. Factual · 18 Apr 2024 AI capabilities
Own records only Factual 18 Apr 2024 Meta's ratio of inference compute to training compute is probably much higher than most other AI companies. Factual · 18 Apr 2024 AI compute, hardware and data
Unresolvable Factual 18 Apr 2024 Feeding the Llama 3 model more tokens would probably have made it somewhat better. Factual · 18 Apr 2024 AI capabilities
Resolved true Factual 18 Apr 2024 Meta started Facebook AI Research (FAIR) about 10 years before 2024. Factual · 18 Apr 2024 Their own work or organisation
Mostly true Factual 18 Apr 2024 Meta AI now generates high-quality images fast enough to generate and update them in real time as the user types. Factual · 18 Apr 2024 AI capabilities
Mixed Factual 18 Apr 2024 Meta's whole GPU fleet is the roughly 350,000 GPUs the interviewer cited as planned by end of year. Factual · 18 Apr 2024 AI compute, hardware and data
Own records only Factual 18 Apr 2024 Meta believes Meta AI is now the most intelligent, freely available AI assistant people can use. Factual · 18 Apr 2024 AI capabilities
Not researched Factual 18 Apr 2024 After the industry standardized on Meta's open-sourced designs, volumes rose, costs fell for everyone, and Meta saved billions of dollars. Factual · 18 Apr 2024 Economics and markets
Not researched Factual 18 Apr 2024 Zuckerberg has not committed to open sourcing everything Meta develops, though Meta is strongly pro open source. Factual · 18 Apr 2024 AI policy and governance
Not researched Factual 18 Apr 2024 Meta is releasing Llama 3 as open source and Llama 3 now powers Meta AI. Factual · 18 Apr 2024 AI capabilities
Not researched Factual 18 Apr 2024 Getting energy permitted is a very heavily regulated government function. Factual · 18 Apr 2024 AI policy and governance
Not researched Factual 18 Apr 2024 In the last few years, GPU production supply constraints limited how many GPUs companies could get, even those with money to pay. Factual · 18 Apr 2024 AI compute, hardware and data
Own records only Factual 18 Apr 2024 Meta spends a lot of time trying to make sure Llama-2 does not help people commit violent acts or things like that. Factual · 18 Apr 2024 AI safety and alignment
Not researched Factual 18 Apr 2024 As of April 2024, GPU supply constraints are easing. Factual · 18 Apr 2024 AI compute, hardware and data
Own records only Factual 18 Apr 2024 There are about 200 million creators on Meta's platforms. Factual · 18 Apr 2024 Economics and markets
Not researched Factual 18 Apr 2024 Picasso said that all children are artists and the challenge is to remain an artist as you grow up. Factual · 18 Apr 2024 Other
Resolved true Factual 18 Apr 2024 Meta did not prioritize coding when working on Llama 2. Factual · 18 Apr 2024 AI capabilities
Not researched Factual 18 Apr 2024 In April 2024, Zuckerberg said he did not think anyone had yet built a gigawatt single AI training cluster. Factual · 18 Apr 2024 AI compute, hardware and data
Not researched Factual 18 Apr 2024 The Llama 3 405B model, still in training, is already at around 85 MMLU. Factual · 18 Apr 2024 AI capabilities
Not researched Factual 18 Apr 2024 Llama 3 70B scores around 82 MMLU and has leading scores on math and reasoning. Factual · 18 Apr 2024 AI capabilities
Not researched Factual 18 Apr 2024 Meta has a long history of open sourcing software. Factual · 18 Apr 2024 Technology
Not researched Factual 18 Apr 2024 At the time being discussed, people understood peace as the temporary time between enemy attacks. Factual · 18 Apr 2024 Other
Not researched Factual 18 Apr 2024 Meta does not tend to open source its products, such as Instagram's code, but open sources low-level infrastructure. Factual · 18 Apr 2024 Technology
Not researched Factual 18 Apr 2024 Augustus became Roman emperor and sought to establish peace. Factual · 18 Apr 2024 Other
Not researched Factual 18 Apr 2024 Zuckerberg said the Open Compute Project was probably the biggest open-source effort in Meta's history, describing it as open sourcing the designs for its servers, network switches and data centers. Factual · 18 Apr 2024 Technology
Unresolvable Factual 18 Apr 2024 Meta has identified about 18 or 19 categories of harmful activity on social media and built AI systems to detect them. Factual · 18 Apr 2024 AI safety and alignment
Not researched Factual 18 Apr 2024 The figure being discussed, who is not named in the quote, sought to change the economy from mercenary and militaristic to positive-sum, which the speaker called a novel idea at the time. Factual · 18 Apr 2024 Other
Own records only Factual 28 Sep 2023 Llama 2 has received more downloads and usage than Zuckerberg expected, even though he was optimistic. Factual · 28 Sep 2023 Their own work or organisation
Unresolvable Factual 28 Sep 2023 Meta has progressed on hand recognition further and faster than Zuckerberg expected. Factual · 28 Sep 2023 Technology
Own records only Factual 28 Sep 2023 As of September 2023, Meta was not close to open sourcing its next foundation model that month and was still somewhat early in working on it. Factual · 28 Sep 2023 Their own work or organisation
Not researched Factual 28 Sep 2023 Zuckerberg said he thinks the mixed reality on Quest 3 is in some ways better than on Quest Pro. Factual · 28 Sep 2023 Technology
Not researched Factual 28 Sep 2023 The headset he describes is 40% thinner, which he says makes it more comfortable; the quoted passage does not say which headset it is being compared with. Factual · 28 Sep 2023 Technology
Not researched Factual 28 Sep 2023 Zuckerberg started working on social networks when they were primarily text; the first Facebook profile had one photo and lists of interests. Factual · 28 Sep 2023 Their own work or organisation
Mostly false Factual 28 Sep 2023 Multiple studies find that most communication, even in speech, is not the words spoken but expression and similar cues. Factual · 28 Sep 2023 Science and research methods
Not researched Factual 28 Sep 2023 Quest 3 does all the VR functions that Quest 2 and other Meta headsets do. Factual · 28 Sep 2023 Technology
Mostly true Factual 28 Sep 2023 Quest 3 runs VR better than Quest 2 because its display and chip are better. Factual · 28 Sep 2023 Technology
Not researched Factual 28 Sep 2023 Meta has an AI dungeon master character embodied by Snoop Dogg. Factual · 28 Sep 2023 Their own work or organisation
Not researched Factual 28 Sep 2023 Zuckerberg said that, in a way, Quest 3 has a stronger chipset than the Pro line at a third of the cost. Factual · 28 Sep 2023 Technology
Not researched Factual 28 Sep 2023 Meta's Kodak Avatar scanning has so far been done by a small number of people in very detailed scans. Factual · 28 Sep 2023 Technology
Resolved true Factual 28 Sep 2023 Zuckerberg often receives criticism for having a relatively stiff facial expression. Factual · 28 Sep 2023 Other
Not researched Factual 28 Sep 2023 Zuckerberg and Lex Fridman both did scans for a Meta research project called Kodak Avatars. Factual · 28 Sep 2023 Technology
Not researched Factual 28 Sep 2023 As of September 2023, Meta is not ready to release a version of its AI that lets a creator build an AI version of themselves. Factual · 28 Sep 2023 Their own work or organisation
Unresolvable Factual 28 Sep 2023 Meta is always training another model beyond its current released one. Factual · 28 Sep 2023 Their own work or organisation
Mostly true Factual 28 Sep 2023 Social networks are now, roughly speaking, in a period where most content is video. Factual · 28 Sep 2023 Technology
Unresolvable Factual 28 Sep 2023 Meta made AI characters based on Jane Austen and Marcus Aurelius. Factual · 28 Sep 2023 Their own work or organisation
Not researched Factual 28 Sep 2023 Zuckerberg said he thinks people's physical bodies also fluctuate and change over time, raising it as part of a question about identity. Factual · 28 Sep 2023 Philosophy, ethics and society
Own records only Factual 28 Sep 2023 In Meta's internal experiments, people had no issues blending photorealistic and non-photorealistic elements in the same scene. Factual · 28 Sep 2023 Technology
Not researched Factual 28 Sep 2023 Meta spent a lot of time on a rigorous assessment and red teaming of Llama 2 before open sourcing it. Factual · 28 Sep 2023 AI safety and alignment
Unresolvable Factual 28 Sep 2023 Meta's Kodak Avatars are much more bandwidth efficient than transmitting full video or 3D immersive video of a scene. Factual · 28 Sep 2023 Technology
Not researched Factual 28 Sep 2023 Llama 2 by itself is not a consumer product but a piece of infrastructure people can build things with. Factual · 28 Sep 2023 Their own work or organisation
Not researched Factual 28 Sep 2023 Meta shipped the Quest Pro headset in 2022 at a price of $1,500. Factual · 28 Sep 2023 Technology
Not researched Factual 24 Mar 2022 VR motion sickness is not fully solved but is getting better with each headset generation. Factual · 24 Mar 2022 Technology
Unresolvable Factual 24 Mar 2022 Apple has a billion iPhones in use. Factual · 24 Mar 2022 Technology
Own records only Factual 24 Mar 2022 Zuckerberg said that a 15-year project he described has six or seven key unknowns, and that multiple teams are trying different approaches to them. Factual · 24 Mar 2022 Their own work or organisation
Not researched Factual 24 Mar 2022 This podcast is the first podcast or public appearance where Zuckerberg has discussed any of Meta's values. Factual · 24 Mar 2022 Their own work or organisation
Not researched Factual 24 Mar 2022 Quest removed the wire and runs VR on a mobile chip with roughly one fiftieth to one one-hundredth the compute power of a powerful desktop. Factual · 24 Mar 2022 AI compute, hardware and data
Mostly true Factual 24 Mar 2022 The original virtual reality devices that Zuckerberg referred to as the ones "we had" did not have hand presence. Factual · 24 Mar 2022 Technology
Mostly true Factual 24 Mar 2022 Heat dissipation is one of the biggest limiting factors for processors running in smart glasses. Factual · 24 Mar 2022 AI compute, hardware and data
Own records only Factual 24 Mar 2022 Each of Meta's company values is coupled with an operational effort. Factual · 24 Mar 2022 Their own work or organisation
Not researched Factual 24 Mar 2022 Distributed work is currently rising. Factual · 24 Mar 2022 AI, the economy and work
Not researched Factual 24 Mar 2022 As Facebook grew, it produced so many bugs that fixing them slowed it down more than moving fast sped it up. Factual · 24 Mar 2022 Their own work or organisation
Not researched Factual 24 Mar 2022 The Oversight Board is a completely independent organization from Meta. Factual · 24 Mar 2022 Their own work or organisation
Mixed Factual 24 Mar 2022 A lot of research shows that people's social, economic and other opportunities are generally pretty anchored to where they physically are. Factual · 24 Mar 2022 Economics and markets
Unresolvable Factual 24 Mar 2022 Until the last few years, Meta had not focused as much on building awe-inspiring things, as opposed to widely used products. Factual · 24 Mar 2022 Their own work or organisation
Unresolvable Factual 24 Mar 2022 Mark Zuckerberg said that Tobi Lütke, founder and CEO of Shopify, recommended to him the book under discussion. Factual · 24 Mar 2022 Other
Not researched Factual 24 Mar 2022 Since 2016, there has almost never been a month in which net public sentiment toward Facebook/Meta was positive. Factual · 24 Mar 2022 Their own work or organisation
Resolved true Factual 24 Mar 2022 When Mark Zuckerberg went to college, he planned to major in classics. Factual · 24 Mar 2022 Other
Not researched Factual 24 Mar 2022 Meta's last company value is "Meta, Metamates, Me", adapted from the naval saying "Ship, shipmates, self." Factual · 24 Mar 2022 Their own work or organisation
Unresolvable Factual 24 Mar 2022 Mark Zuckerberg says no interviewer had previously started an interview by asking him about fencing. Factual · 24 Mar 2022 Other
Not researched Factual 24 Mar 2022 A hydrofoil board has almost no friction compared to a surfboard. Factual · 24 Mar 2022 Other
Back to the list

Factual Their own work or organisation

“We’re building teams internally to enable as many developers and partners as possible to use Llama, and we’re actively building partnerships so that more companies in the ecosystem can offer unique functionality to their customers as well.”

Mark Zuckerberg, Meta CEO, writing on Meta Newsroom 23 July 2024 Open Source AI is the Path Forward (Meta Newsroom)

Meta is building internal teams to enable developers and partners to use Llama and is actively building partnerships with ecosystem companies.

Mostly true True in substance, wrong in a detail.

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

What would count as true or false

Judged as of 2024-07-23. The claim has two parts: (1) Meta was building internal teams to help developers and partners use Llama, and (2) Meta was actively building partnerships with ecosystem companies. True: both are documented around that date. For (2), this means announced Llama partners such as AWS, Microsoft Azure, Google Cloud, NVIDIA, Databricks, Groq, Dell or Snowflake. For (1), this means evidence of internal teams or programs supporting Llama adoption. Mostly true: the partnerships are well documented and the internal teams are only lightly evidenced. Mixed: only one part is supported. Mostly false: there is little evidence for either part beyond the statement itself. False: evidence shows Meta was not doing either.

Evidence

  1. Meta releases the biggest and best open-source AI model yet The Verge · 23 July 2024
    “To help get Llama 3.1 out into the world, Meta is working with more than two dozen companies, including Microsoft, Amazon, Google, Nvidia, and Databricks, to help developers deploy their own versions.”
  2. Why Meta’s Llama 3.1 is a boon for enterprises and a bane for other LLM vendors InfoWorld · 24 July 2024
    “Meta said it had partnered with the likes of Accenture, AWS, AMD, Anyscale, Cloudflare, Databricks, Dell, Deloitte, Fireworks.ai, Google Cloud, Groq, Hugging Face, IBM watsonx, Infosys, Intel, Kaggle, Microsoft Azure, Nvidia DGX Cloud, OctoAI, Oracle Cloud, PwC, Replicate, Sarvam AI, Scale.AI, SNCF, Snowflake, Together AI, and the UC Berkeley vLLM Project”
  3. Introducing Llama 3.1: Our most capable models to date Meta AI (ai.meta.com) Own record
    “We’d like to thank and acknowledge (in alphabetical order): Accenture, Amazon Web Services, AMD, Anyscale, CloudFlare, Databricks, Dell, Deloitte, Fireworks.ai, Google Cloud, Groq, Hugging Face, IBM WatsonX, Infosys, Intel, Kaggle, Microsoft Azure, NVIDIA, OctoAI, Oracle Cloud, PwC, Replicate, Sarvam AI, Scale.AI, SNCF, Snowflake, Together AI, and vLLM project developed in Sky Computing Lab at UC Berkeley.”
  4. Meta’s Llama models get 350 million downloads Computerworld · 30 August 2024
    “We’ve grown the number of partners in our Llama early access program by 5x with Llama 3.1 and will do more to meet the surging demand from partners”

Reasoning

Independent launch-day reporting confirms the partnership part. E1 says Meta was working with more than two dozen companies, including Microsoft, Amazon, Google, Nvidia and Databricks, to help developers deploy Llama 3.1. E2 lists named partners such as AWS, Azure, Google Cloud, Groq, Dell and Snowflake, and Meta's own acknowledgement list in E3 matches it. The internal-teams part has only light support: E4 quotes Meta saying the number of partners in its Llama early access program grew fivefold with Llama 3.1, but nothing directly documents internal teams set up for this purpose.

Cite this claim

Mark Zuckerberg, “Meta is building internal teams to enable developers and partners to use Llama and is actively building partnerships with ecosystem companies.” (said 23 July 2024). Verdict: mostly true. Reliability API, claim c_050bf2e98cf2, method 0.1.0+3f84d14d. https://epistacklabs.com/claims/c_050bf2e98cf2

Claims without a verdict

Is a verdict here wrong, or do you speak for Mark Zuckerberg? Reply on the claim's own page, with the form at its foot (how replies work). Verdicts are made automatically; how verdicts and scores are made.