MosaicML·released 2023-06-221 source
MPT-30B benchmark scores: 8 benchmarks tracked. 0% of its scores are independently reproduced — the rest are self-reported or unverified.
Leads on 0 of 8 benchmarks·best result 73.6 on TriviaQA (unverified)·0 of 8 independently reproduced
| Benchmark | Score | Evidence status | Measured |
|---|---|---|---|
| TriviaQA | 73.6accuracy (%) | unverified· optimizedT1 | 2023-06-22Llama 2: Open Foundation and Fine-Tuned Chat Models |
| PIQA | 63.8accuracy (%) | unverified· optimizedT1 | 2023-06-22Llama 2: Open Foundation and Fine-Tuned Chat Models |
| Winogrande | 42accuracy (%) | unverified· optimizedT1 | 2023-06-22Llama 2: Open Foundation and Fine-Tuned Chat Models |
| OpenBookQA | 36accuracy (%) | unverified· optimizedT1 | 2023-06-22Llama 2: Open Foundation and Fine-Tuned Chat Models |
| GSM8K | 34.4accuracy (%) | unverified· optimizedT1 | 2023-06-22Stanford HELM |
| ARC AI2 | 34.13accuracy (%) | unverified· optimizedT1 | 2023-06-22Llama 2: Open Foundation and Fine-Tuned Chat Models |
| MMLU | 30.53accuracy (%) | self-reported· optimizedT1 | 2023-06-22Qwen Technical Report |
| BBH | 17.33accuracy (%) | self-reported· optimizedT1 | 2023-06-22Qwen Technical Report |
Claims drawn from cited facts, not live model generation.
Developed by MosaicML, this model was released in 2023. With 30 billion parameters, it is a large model whose scale provides considerable capability headroom at the expense of higher computational cost.
3 cited facts
This model is tracked across eight benchmarks and holds no current top score. Because these results are self-reported, they are vendor-claimed and have not been independently confirmed; read them as claims rather than verified results.
3 cited facts
With an average gap of 44.1 points under the disclosed harness, this model trails benchmark leaders by a wide margin and reads as well behind the front of the pack. It holds the top score on none of the tracked benchmarks, so it shows no genuine category leadership yet.
2 cited facts
2 facts cross-checked across data sources: 2 single-source.
Source facts, citations, and refresh stamp for this record.
Sources: curated_param_counts, epoch_benchmarks
MPT-30B is an AI model developed by MosaicML, released 2023-06-22. Its benchmark record below tags every score as independently reproduced, self-reported, or unverified.
MPT-30B has recorded scores on 8 benchmarks, each shown with its evidence status.
MPT-30B has recorded scores on 8 benchmarks — PIQA, OpenBookQA, Winogrande, TriviaQA, MMLU, ARC AI2, and 2 more. The full table above shows each score with its evidence status.
0 of 8 recorded scores (0%) are independently reproduced rather than self-reported by the lab.
MPT-30B has 8 tracked claims: 2 self-reported, 6 unverified.