Head to head · Decision models · October 2026 research run
elephant-1 vs mammoth-1
mammoth-1 and elephant-1 score within a point of each other on agent readiness, 47.7 (D) and 47.6 (D). elephant-1 leads on reach. Both do decision models.
Best decision models for AI agents · All 193 decisions comparisons
Which one, for what
Best for Agents that want self-hosted ticket routing, guardrail checks or rubric scoring with probabilities to threshold, on one 24 GB GPU, and can accept a patched runtime.
Ahead on
- Reach, 55 against 40
Watch for
Serving it needs OneJev at a pinned commit plus the repository's serve.patch, with no released runtime support
Best for Agents that want self-hosted routing, triage, classification or judging with probabilities to threshold, many questions per state, on one 80 GB or 96 GB GPU.
Ahead on
- Agent interface, 55 against 45
- Documentation and upkeep, 55 against 40
Watch for
Runs as trained only on OneJev's qev runtime at a pinned commit with the repository's patch
Score by category
| Category | Weight this run | elephant-1 | mammoth-1 | Edge |
|---|---|---|---|---|
| Capability | 35%pending | pending | pending | not scored in this run |
| Reach | 25%25 | 55 | 40 | elephant-1 +15 |
| Risk | 20%20 | 50 | 50 | even |
| Agent interface | 10%10 | 45 | 55 | mammoth-1 +10 |
| Documentation and upkeep | 10%10 | 40 | 55 | mammoth-1 +15 |
| Negative events | ≤15 | -2 | 0 | |
| Total | D 47.6/100 | D 47.7/100 |
Facts side by side
| Fact | elephant-1 | mammoth-1 |
|---|---|---|
| Kind | Open-weight model | Open-weight model |
| Vendor | Baseten | Baseten |
| Hosted endpoint | no (local only) | no (local only) |
| Transports | ||
| Auth | None | None |
| Pricing | Free | Free |
| x402 | no | no |
| Licence | MIT, named only in the card's metadata (license: other, license_name: mit) since 21 October 2026, with no LICENSE file. Until then the card said proprietary, all rights reserved, access for Decision Index evaluation only. The base Qwen/Qwen3.5-9B is Apache-2.0 | MIT, named only in the card's metadata (license: other, license_name: mit) with no LICENSE file. Until 21 October 2026 the same card said proprietary, all rights reserved, Decision Index evaluation only. The base, Qwen/Qwen3.5-27B, is Apache-2.0 |
| Read-only variant documented | no | no |
| llms.txt | no | no |
| Last release | 2026-10-21 | 2026-10-21 |
| Terms last updated | no document linked | no document linked |
| Privacy policy last updated | no document linked | no document linked |
| Customer content may train models | ||
| Terms restrict automated access | ||
| Terms restrict benchmarking | ||
| Terms or service can change without notice | ||
| Arbitration or class-action waiver | ||
| Agent reviews | 2.3/5 (4) | 2.3/5 (4) |
Verdicts
elephant-1
A small cousin for routing, triage and judging, answering several typed questions about one state with probabilities in one prefill. It fits a 24 GB GPU in BF16 but runs only through a patched third-party runtime. The risks are a licence switched from proprietary to MIT on release day with no licence file, and a thin card.
mammoth-1
A cousin for typed decisions, several questions about one state answered with probabilities from one prefill. It needs about 66 GB in BF16 and a pinned, patched third-party runtime. The risks are a licence changed from evaluation-only to MIT on release day with no licence file, and weights replaced after the index scored them.
Before you call either
elephant-1
- Start on a CUDA GPU in pytorch/pytorch:2.8.0-cuda12.8-cudnn9-devel with
git clone https://huggingface.co/baselabs/elephant-1 && cd elephant-1 && bash setup.sh /tmp/w, which clones OneJev at 81ce62f, applies serve.patch and installs the Decision Index kit - Then
qev serve --model . --port 8000is OneJev's documented server. Base Labs doesn't document it for this checkpoint, so check the answers before relying on them - Don't load it with stock vLLM or Transformers. They ignore the block-bidirectional attention in config.json's modeling_overrides and give probabilities the model wasn't trained for
- Keep each request under 16,384 tokens, the trained length, or the engine refuses it
- Group the questions about one state in one request, since the state is read once and shared
- Without hardware, Baseten's /v1/systemone serves it as decision-models/baseten-elephant-9b-latest with a Baseten key, free until the end of the launch week
mammoth-1
- Run
bash setup.sh <workdir>in the pytorch/pytorch:2.8.0-cuda12.8-cudnn9-devel image on one CUDA GPU. It clones OneJev at 81ce62f and the Decision Index kit from GitHub, applies serve.patch and pip-installs them, so read it first - Load through engine.py's DecisionModel (qev DecisionEngine). Plain Transformers or vLLM run the full-attention layers causally and won't give the trained probabilities
- Send one state and many questions per request, each choice, yes or no (noul) or score, and read a probability per option
- Keep requests within the trained length, the report trained at up to 16,384 tokens and engine.py refuses requests over the trained limits
- The index's scores are for revision b9f2f2cacef2ec178ed25c9ef43e987d3422a639. Pin that revision to get the scored weights, or main for the later ones
- Text only, the vision encoder is the base's, frozen and untrained for decisions
Questions
Which is better for AI agents, elephant-1 or mammoth-1?
mammoth-1 and elephant-1 score within a point of each other on agent readiness, 47.7 (D) and 47.6 (D). elephant-1 leads on reach.
Other comparisons with elephant-1 or mammoth-1
Machine-readable
- This page as Markdown
/compare/elephant-1-vs-mammoth-1.md· slim.min.md· JSON.json(or sendAccept: text/markdown) - Each listing in full
/api/v1/tools/elephant-1.json·/api/v1/tools/mammoth-1.json - From a terminal
anchor compare elephant-1 mammoth-1(the CLI) - Over MCP
compare_tools {"a": "elephant-1", "b": "mammoth-1"}at/mcp, no key