AI supercomputer
MPT-30B training cluster
MPT-30B training cluster is an AI supercomputer, with 256 H100e of compute, owned by MosaicML.
Facts
Every value links to its source, and to the source's own record where it has one. Licences and last-checked dates are per row.
| Fact | Value | Source |
|---|---|---|
| Status | Operating | Epoch AI supercomputers · CC BY 4.0 · checked 24 Oct 2026 |
| Start year | 2023 | Epoch AI supercomputers · CC BY 4.0 · checked 24 Oct 2026 |
| Owner | MosaicML | Epoch AI supercomputers · CC BY 4.0 · checked 24 Oct 2026 |
| Compute (H100 equivalents) | 256 H100e | Epoch AI supercomputers · CC BY 4.0 · checked 24 Oct 2026 |
| Chips | NVIDIA H100 SXM5 80GB | Epoch AI supercomputers · CC BY 4.0 · checked 24 Oct 2026 |
| AI chips | 256 | Epoch AI supercomputers · CC BY 4.0 · checked 24 Oct 2026 |
| Facility power | 0.4 MW (estimated) | Epoch AI supercomputers · CC BY 4.0 · checked 24 Oct 2026 |
Research on this AI supercomputer and its owner.Who owns it up to the parent company, its permits and status, and an introduction to the owner.
Request researchHow it compares by compute
MPT-30B training cluster is the 162nd-largest of the 284 AI supercomputers in our data by compute. Ranks and totals cover the AI supercomputers in our data that report their compute, not every one in existence.
| Within | Rank |
|---|---|
| All AI supercomputers in our data | 162nd of 284 |
Owner and place
More from MosaicML
- MosaicML MPT training cluster139 H100e
Data on this page
- Epoch AI. Epoch AI supercomputers, CC BY 4.0.