| --- |
| pretty_name: ClassOne Gemma 4 E4B — Decision Index 0.3 public run |
| tags: |
| - decision-index |
| --- |
| |
| # ClassOne Gemma 4 E4B — Decision Index 0.3 public run |
|
|
| Complete, unmodified results of the public Decision Index 0.3 suite for |
| [devops-thiago/classone-gemma4-e4b](https://huggingface.co/devops-thiago/classone-gemma4-e4b) @ `b4f6cf1`, |
| served with [devops-thiago/class-one](https://github.com/devops-thiago/class-one) v0.2.3 |
| (`d1852461c33ad805d7212ebc3634fa4ce96cf0ab`) and run with the |
| [apolinario/decision-index](https://github.com/apolinario/decision-index) kit at `25c6ec7` |
| (stock `http` engine, `pipeline --compact --upload`). |
|
|
| - Decision Index (public 0.3): **26.4** (raw 43.49, breadth 24.39); `complete: true`, 140,178 of 140,178 scored requests answered, all `ok`. |
| - Hardware: 1x NVIDIA RTX PRO 6000 Blackwell Server Edition (96 GB), one request at a time, 17,955.6 s. |
| - The run also executed the edition's 442 excluded rows. 12 of those (8 ToolRet chunks, 4 BRIGHT chunks) returned HTTP 500 Internal Server Error; they are all edition exclusions (`beyond_nearly_every_context_window` / sibling chunks) and are not part of the scored set. 0 unsupported, 0 errors among the 140,178 scoreable requests. |
|
|
| Files in `runs/classone-gemma4-e4b-v0.2.3/`: `results.jsonl.gz`, `scores.json`, `index.json`, |
| `benchmark-summary.json`, `environment.json`, `status.json`, exactly as written by the kit. |
| `results.jsonl.gz` is compact (no request payloads or raw outputs); the suite itself is not included. |
| `results.jsonl.gz` sha256 `501f4268b91cd9b8d47084013a8880227d1145f25c153171ef5a02d423a33704`. |
|
|
