# LLM Pretraining on 500-shard corpus (val)

Source: [LLM Pretraining on 500-shard corpus (val)](<https://www.sota2.com/research/sota/llm-pretraining-on-500-shard-corpus-val>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-04-13T07:39:38.523Z

Evaluation of Large Language Model pretraining performance using the validation split of the 500-shard corpus, typically measured by Bits Per Byte (BPB).

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

| Scope | Recorded value |
| --- | --- |
| Task | LLM Pretraining |
| Dataset (including any stated split) | 500-shard corpus (val) |
| Metric | Validation BPB |
| Unit | SCALAR |
| Direction | Lower is better |
| Results loaded | 3 of 3 |
| Coverage | All results currently returned for this benchmark |

Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol.

## Metrics

- [Validation BPB](<https://www.sota2.com/research/sota/llm-pretraining-on-500-shard-corpus-val.md?metric=Validation%20BPB>): Lower is better; unit: SCALAR.

## Recorded results

| Method | Validation BPB | Result date | Report paper |
| --- | --- | --- | --- |
| Opus | 0.758 | Not recorded | 69dc788a7d0172140ec78aa3 |
| Sonnet | 0.869 | Not recorded | 69dc788a7d0172140ec78aa3 |
| GPT-5.2 | 0.97 | Not recorded | 69dc788a7d0172140ec78aa3 |

### Opus

Validation BPB: 0.758. Result ID: 69dc9d8447f34ce924c34c95.

[Reporting paper](<https://www.sota2.com/research/paper/69dc788a7d0172140ec78aa3.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69dc788a7d0172140ec78aa3.md>)

Recorded evaluation configuration: {"vs. baseline":"−22%","Cost":"$200"}. This record does not assert independent reproduction.

### Sonnet

Validation BPB: 0.869. Result ID: 69dc9d8447f34ce924c34c94.

[Reporting paper](<https://www.sota2.com/research/paper/69dc788a7d0172140ec78aa3.md>)

Recorded evaluation configuration: {"vs. baseline":"−10%","Cost":"$120"}. This record does not assert independent reproduction.

### GPT-5.2

Validation BPB: 0.97. Result ID: 69dc9d8447f34ce924c34c93.

[Reporting paper](<https://www.sota2.com/research/paper/69dc788a7d0172140ec78aa3.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69dc788a7d0172140ec78aa3.md>)

Recorded evaluation configuration: {"Cost":"$150"}. This record does not assert independent reproduction.

Page 1 of 1. This is an index page, not the full collection.
