ClaimsMarc Andreessen

“It's completely open - source, and anyone can download and run it.”

Marc Andreessen, a16z co-founder 7 January 2026 2026 Early Interview with Marc Andreessen: AI Revolution Just Started (36Kr) archived copy

Factual AI capabilities

The DeepSeek model is completely open-source and anyone can download and run it.

Mostly true True in substance, wrong in a detail.

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

What would count as true or false

Judged as of 2026-01-07 for the DeepSeek model under discussion (for example DeepSeek-R1 or V3). True: the model weights are publicly released under a permissive or open licence (such as MIT) and can be freely downloaded (for example from Hugging Face) and run locally by anyone with suitable hardware. Mostly true: the weights are openly downloadable and runnable, but 'completely open-source' slightly overstates it because training data or full training code were not released (open weights rather than fully open source). Mixed: the weights are downloadable only under significant licence restrictions, or only some variants are open. Mostly false: only limited or distilled versions are downloadable. False: the weights are not publicly downloadable, and access is only by API.

Set before the research began.

Evidence (4)

  1. E1 DeepSeek-R1 Release DeepSeek · 20 January 2025

    “Code and models are released under the MIT License: Distill & commercialize freely!”
  2. E2 DeepSeek-R1/README.md at main · deepseek-ai/DeepSeek-R1 GitHub (deepseek-ai/DeepSeek-R1 repository) · 20 January 2025

    “This code repository and the model weights are licensed under the MIT License.”
  3. E3 Understanding the DeepSeek model license: Balancing openness and responsibility Black Duck · 4 February 2025

    “Although the source code for DeepSeek is licensed under the permissive MIT license, the DeepSeek model itself is governed by the DeepSeek license, which does not meet the definition of an open source license.”
  4. E4 Open Weights vs Open Source: The Real Difference (2026) geotoolbox.ai · 3 July 2026

    “But DeepSeek does not release its training data or full training code, so you cannot reproduce or fully audit the model.”

Reasoning

DeepSeek's R1 release announcement and the R1 repository README both say the model weights are released under the MIT License (E1, E2). That meets the criteria for openly downloadable, permissively licensed weights. However, E4 says DeepSeek does not release its training data or full training code, so "completely open-source" overstates it: these are open weights, not fully open source. E3 says the DeepSeek model itself is governed by a DeepSeek licence that does not meet the definition of an open-source licence, while E2 says the R1 model weights are licensed under MIT; the quoted passage from E3 does not specify which model it refers to.

Cite this claim

Marc Andreessen, “The DeepSeek model is completely open-source and anyone can download and run it.” (said 7 January 2026). Verdict: mostly true. Reliability API, claim c_0760c435ee7e, method 0.1.0+c876030c, dataset release v1.1.2. https://epistacklabs.com/claims/c_0760c435ee7e

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

Reply to this claim

For Marc Andreessen, 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 Marc Andreessen

  1. Mixed Factual 7 January 2026
    “Basically, within less than 12 months, four or five Chinese companies have caught up to the industry front - line.”

    Within less than 12 months, four or five Chinese companies caught up to the frontier of the AI industry.

  2. Mostly true Factual 7 January 2026
    “The idea of neural networks was explored by a small group of people for 80 years.”

    Neural networks were explored by a small group of people for about 80 years.

  3. Mostly true Factual 7 January 2026
    “When I was in college in the 1980s, AI was a niche field, and everyone thought it would never come true.”

    In the 1980s, when Andreessen was in college, AI was a niche field and everyone thought it would never come true.

  4. Resolved true Factual 7 January 2026
    “If you track the capabilities of cutting - edge models over time, you'll find that after 6 or 12 months, there will be a small model with the same capabilities.”

    Within 6 to 12 months of a frontier model's release, a small model with the same capabilities appears.

  5. Mostly false Factual 8 October 2025
    ““ChatGPT is in fact a better doctor than your doctor today, with almost a hundred percent certainty,””

    ChatGPT is today a better doctor than your doctor, with almost a hundred percent certainty.

All of them

Verdicts are made automatically; how verdicts are made.