Ctrl K
Contact
Support · Meta Llama

Will it last?

Ownership levelPartialnone·limited·partial·substantial·fullAnalytical input B · 72.4/100

This page is a projection of the one entry record, the Transparency factor that Support covers. The full verdict is set by all four factors together, floor-weighted so the weakest caps the whole.

Which domain expands which factor
  • AssessUse & modify + Transparency
  • ImplementData control + Doesn't fail you
  • UseDoesn't fail you
  • SupportTransparency

Common problems & fixes

The two most frequent problems are operational, not model quality:

  • Gated-access failures (401/403, "you do not have access"). You must accept the license on the model page and authenticate with a Hugging Face token that has been granted access. Approval is not always instant. Use the same account for acceptance and token.
  • Tokenizer / chat-template mismatches. Hand-built prompt strings or the wrong special tokens produce rambling, role-confused, or never-terminating output. Always render prompts with the tokenizer's apply_chat_template and make sure the end-of-turn token is configured as a stop token in your serving stack. Quantized community builds occasionally ship a stale tokenizer/template - cross-check against the canonical meta-llama repo.

Versions & changelog

Meta ships on a predictable generational cadence: Llama 3.1 (8B/70B/405B) → Llama 3.3 (70B refresh narrowing the gap to 405B) → Llama 4 (Scout/Maverick/Behemoth, mixture-of- experts, multimodal). Each generation is released with model cards documenting the changes. There is no rolling per-commit changelog for the weights; the durable way to track and pin a specific build is the Hugging Face revision (commit hash) on the canonical org. Always pin a revision in production rather than tracking main.

Security disclosure process

Meta provides a Responsible Use reporting path for model issues, policy violations, and feedback (via the Responsible Use Guide and associated reporting channels). What is not present is a formal, CVE-style vulnerability-disclosure process or cryptographic signing for the model weights themselves - this is the gap noted in the governance score. Coverage is therefore partial: there is a documented way to report model behaviour, but no published security-response SLA or signed-artifact verification path.

Community & support channels

Support is community- and platform-based rather than a formal support contract: the Meta Llama community resources and documentation on llama.com, Hugging Face model discussions on each repo, and Meta's GitHub repositories (llama-models, llama-recipes/ cookbook, torchtune) for issues and examples. The broad serving-ecosystem projects (vLLM, llama.cpp, Ollama) also carry active Llama-specific issue trackers.

Tracked known issues

The model cards document the standing known issues honestly: hallucination in the class-standard profile, residual jailbreak and prompt-injection susceptibility, and the explicit warning that base checkpoints are not safety-tuned and must not be deployed as chat models. Long-context recall softens mid-window, and multilingual quality outside the eight officially supported languages is not validated. None are regressions unique to a release; they are the persistent caveats the safe-deployment controls are designed to contain.

How this scores

The ownership factor this domain covers, drawn from the one entry record.

2

TransparencyDo you know what it is: weights, training, behaviour, and legible terms?

Moderate

You can inspect and self-guard the weights, and Meta publishes safety evaluations plus Llama Guard and Prompt Guard, but training data and code are closed and no broad independent red-team spans every harm domain - you see the model, not how it was made.

How this scores (AOI sub-dimensions)
Provenance4/5how well we can trace and verify what went into the modelVerified Meta org on Hugging Face, safetensors-only distribution, published checksums, clear canonical source, and no malicious-checkpoint incident on the canonical org (checklist ~6/8).
Governance4/5how accountable and well-documented the publisher isReputable, legally accountable publisher with a predictable multi-generation release cadence, a Responsible Use Guide and a reporting path.
What this means for adoptionYou own the self-hosted Llama weights in practice: they run on your infrastructure, perform at the top of the open field, and pair with Meta's Llama Guard and Prompt Guard. But overall ownership is partial, not substantial - the Community License is not a clean grant: an access gate, 'Built with Llama' attribution, a 700M-MAU commercial trigger, and no multimodal licence for EU-domiciled entities mean you cannot freely use and modify the whole family the way an OSI licence allows, and training is closed. Clear those conditions first: check the MAU trigger, honour the attribution terms, and if you are EU-domiciled avoid the multimodal variants entirely.

Sources

The same evidence records as the entry sheet. Read means the text was verified; unverified means it is known to exist but not yet read.

Licenceread2026-07-25
Llama 4 Community License, read verbatim: the "Additional Commercial Terms" 700M-MAU threshold ("greater than 700 million monthly active users in the preceding calendar month, you must request a license from Meta"), the "Built with Llama" attribution clause, California governing law, and - in the incorporated AUP - the EU multimodal restriction denying the licence to individuals domiciled in, or companies with a principal place of business in, the European Union.
Terms of serviceread2026-07-25
Llama 4 Acceptable Use Policy, read verbatim: six prohibited-use categories (illegality/rights infringement; death or bodily harm incl.
Third-party analysisread2026-07-25
Llama family variants, EU multimodal licensing restriction, ecosystem/cloud availability, and Meta declining the EU GPAI Code of Practice.
Model cardread2026-07-25
Canonical, verified and gated meta-llama organisation on Hugging Face distributing safetensors with checksums; per-version licence tags (llama4, llama3.3, llama3.2, llama3.1).
Third-party analysisread2026-07-25
Llama is competitive-to-leading among open-weight generalist families on public leaderboards.
Documentationunverified2026-07-25
Meta ships safety-tuned instruct variants plus Llama Guard and Prompt Guard companion classifiers with published safety evaluations and a responsible-use reporting path.
Model cardread2026-07-25
Llama 3.1 Instruct model card on Hugging Face: 128K context window, the Llama 3 chat template and special tokens, the eight officially supported languages, and documented tool-calling.