---
title: mammoth 1 on SIE | Superlinked
description: SIE, an open-source inference engine, serves mammoth 1, an open model that answers typed questions with choices, probabilities and scores.
canonical_url: https://superlinked.com/models/baselabs-mammoth-1
last_updated: 2026-11-20
---

# mammoth 1

An open model that answers typed questions with choices, probabilities and scores.

Model id: `baselabs/mammoth-1` · Endpoint: `extract` · [Hugging Face](https://huggingface.co/baselabs/mammoth-1)

Listed under [Typed decisions](/models#decisions) on [the model index](/models.md).

| Accuracy | Time per request | Serverless price |
|---|---|---|
| 76.3% (Typed Decisions, answers right) | – (Not timed one request at a time) | Not on Serverless (run in your cloud) |

## Next to the other decision models

| Model | Typed Decisions, answers right | Time per request | Serverless price | Size |
|---|---|---|---|---|
| [mammoth 1](/models/baselabs-mammoth-1.md) `baselabs/mammoth-1` | 76.3% | – | Not on Serverless (run in your cloud) | 27.4B |
| [DeBERTa v3 large zero-shot](/models/moritzlaurer-deberta-v3-large-zeroshot-v2-0.md) `MoritzLaurer/deberta-v3-large-zeroshot-v2.0` | – | 13.8 ms per decision | $0.129 per 1M tokens | 435M |
| [KnowLine 4B Gen5.5](/models/pelaai-knowline-4b-gen5-5.md) `PelaAI/KnowLine-4B-Gen5.5` | 78.8% | – | Not on Serverless (run in your cloud) | 4.7B |
| [KnowLine 31B Gen1](/models/pelaai-knowline-31b-gen1.md) `PelaAI/KnowLine-31B-Gen1` | 78.5% | – | Not on Serverless (run in your cloud) | 31.3B |
| [pplx decider v1.2](/models/perplexity-ai-pplx-decider-v1-2.md) `perplexity-ai/pplx-decider-v1.2` | 75.5% | – | Not on Serverless (run in your cloud) | 26.1B |

– = not measured / not timed / not listed.

[All 8 decision models](/models#decisions)

## Measured on public datasets

| Dataset | Metric | Value |
|---|---|---|
| Banking77 ([measured internally](https://superlinked.com/docs/evals/methodology#measured-internally)) | Accuracy | 93.5% |
| XNLI ([measured internally](https://superlinked.com/docs/evals/methodology#measured-internally)) | Accuracy | 93.0% |
| BoolQ ([measured internally](https://superlinked.com/docs/evals/methodology#measured-internally)) | Accuracy | 92.8% |
| AG News (decision) ([measured internally](https://superlinked.com/docs/evals/methodology#measured-internally)) | Accuracy | 88.8% |
| Typed Decisions ([measured internally](https://superlinked.com/docs/evals/methodology#measured-internally)) | Accuracy | 76.3% |
| emotion ([measured internally](https://superlinked.com/docs/evals/methodology#measured-internally)) | Accuracy | 59.0% |
| SST-5 ([measured internally](https://superlinked.com/docs/evals/methodology#measured-internally)) | Accuracy | 55.2% |

## Facts

- Size: 27.4B parameters
- License: MIT
- Reads: Text, up to 8,192 tokens
- Endpoint: extract

## Evaluation

- Accuracy: Run on SIE against the public dataset named in each figure
- Time: SIE timed on a GPU server with the client alongside, one request at a time
- Prices: Serverless list prices, effective 10 Nov 2026
