Llama 3 / Llama 3.1 405B (Meta)

A dated reading of what is claimed, reported, and independently verified in the current evidence.

Current reading

Readiness band and full integer triple

AI-assisted assembly · derived results

Readiness band

Mature

Primary summary from verified readiness

Evidence strength Strong

Decision

Human editorial judgment · 2026-07-23

Low-friction candidate

Why: Meta's Community License terms (>700M MAU clause) don't burden Pragma.Vision's scale, and the model is well-documented and broadly available via major clouds, but current evidence shows Meta itself has moved on to Llama 4 and open-weight competitors have closed the gap, so it is low-friction to reference but not the model to newly adopt for production.
Next: Track Llama 3.1 405B as the historical open-weight reference baseline; if self-hosting is pursued, prefer a current-generation open-weight model (Llama 4, Qwen, DeepSeek, Kimi, GLM — all already tracked) since 405B is two generations behind Meta's own line and heavy to self-host.

Constraints

No named blocker is present in the current public projection.

Evidence summary

AI-assisted assembly

Counts and dates only. Raw signals, private excerpts, trust records, and internal corpus material are not published here.

Publication record

Initial public reading