Exact illustrative checkpoint sizes under a 14 B/parameter serialized-state recipe
2.45–14 TB (exact assumed layout)derived
| Value kind | derived — Derived values are guide calculations; their result is only as fixed as the stated inputs, scope, and method. |
|---|---|
| Scope | Illustrative checkpoint layout using decimal TB; excludes optional gradient copies, sharding metadata and other checkpoint state. |
| As of | 2026-09-08 |
| Source | Guide, exact illustrative parameter counts and decimal-byte arithmetic, 8 September 2026; serialized-state recipe is canonical in Chapter 9.4. |
| Derivation | Exact assumed 175e9 and 1e12 parameters, each serialized as 2 B BF16 weights +4 B FP32 master +4 B+4 B FP32 Adam moments =14 B; products are 2.45e12 and 14e12 B. Excludes gradients and other run state; inspect the real serializer. |
| Review | checking…review by 2026-11-23 · standard cadence |
| Recorded changes | last 2026-09-16 · 5 revisions tracked |
| Claim id | checkpoint-size-for-a-175b-to-1t-param-model-at |
Where the guide uses it
← Full numbers register — every date-stamped figure in the guide, with revision history.