elephant-1

# Fix list: elephant-1

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/elephant-1, the October 2026 research run, assessed 21 October 2026. Grade D, 47.6 out of 100.

This is everything the published grade says the listing lacks, the biggest possible gain to the total first. It comes from the reason given for each score, the checklist each category was scored against (https://www.anchorterminal.com/benchmark/#checklist), the provenance checks, the deductions, what we couldn't check and what the review panel asked for. A fix counts at the next check, once it's public.

For a coding agent working on elephant-1: work through the items below in the product, its docs and its public pages. Each category gives the reason for its score, with the points each checklist item earned, and the checklist itself, so the gap is the items that earned less than their points. Change the product, not the wording, and keep a note of what you changed and where it's published.

## 1. Reach, 55 out of 100, up to 17.3 more on the total

Why it scored 55: Memory 25 of 30, the BF16 repository is 18.8 GB, 22.6 GB with a fifth added, so it fits 24 GB. Runtimes 0, only a third party's OneJev runtime patched by the repository's serve.patch applies the model's changed attention and readout. Builds 0, no quantised build. Terms 15, MIT in the card's metadata with nothing added, though there's no licence file. Download 10, not gated. Hosted 5, Baseten's own Model API serves it.

The checklist (https://www.anchorterminal.com/benchmark/#checklist-reach):

Whether an agent can run it, and where.

- 0 to 30, memory. The smallest build from the publisher or a runtime's own library that keeps 4-bit precision or better, at its file size plus a fifth for the runtime and a working context. 30 for 16 GB or less, 25 for 24 GB, 20 for 32 GB, 15 for 64 GB, 10 for what fits our reference machine (the GPU share of 128 GB of unified memory), 5 for 256 GB, 0 above.
- 0 to 25, runtimes. 5 for each of llama.cpp, Ollama, vLLM, SGLang, LM Studio and Hugging Face Transformers that runs the architecture in a released version, from the runtime's own code, library or documentation. A publisher's fork or an unreleased branch doesn't count. 25 at most.
- 0 to 15, builds from the publisher. 15 for quantised builds at two precisions or more (FP8, 4-bit, GGUF), 10 for one, 0 when only other people's quantisations exist.
- 0 to 15, terms of use. 15 for an OSI-approved licence with nothing added, 10 for a custom licence that lets anyone use it commercially for free below a threshold or under a use policy, 5 for research or non-commercial use only, 0 for no licence or unclear terms.
- 0 to 10, download. 10 when the files download with no account and no terms to accept, 5 for a click-through that needs an account, 0 when a person approves each request.
- 5, hosted too. The publisher's own API serves it, or at least two inference providers do, so an agent without the hardware can still call it.

When our reference-machine runs start, speed on that machine (tokens per second and time to first token at the build that fits) joins Reach.

## 2. Risk, 50 out of 100, up to 15.4 more on the total

Why it scored 50: File safety 15 of 25, safetensors only, but serving it means running the repository's setup.sh and patch. Provenance 15, the organisation Baseten links to, with a checksum on every file. Safety evaluation 0, none for the model. Agent guidance 5, threshold and escalation advice only. Licence risk 10, the licence went from proprietary to MIT on release day with no note. Security programme 5, Baseten's disclosure address but no public advisories.

The checklist (https://www.anchorterminal.com/benchmark/#checklist-risk):

What could go wrong when you run it, and what the publisher has done about it. A higher score means less risk.

- 0 to 25, file safety. 25 for weights in safetensors or GGUF only, 15 when pickle files (.bin, .pt, .pth) sit beside them, 0 for pickle only. 10 off when the documented way to load it or to build its prompts runs code from the repository (`trust_remote_code`, or the publisher's scripts in place of a template).
- 0 to 15, provenance. 10 when the files come from the publisher's own organisation, the one its site or docs link to, and 5 for a checksum on every file.
- 0 to 20, safety evaluation. 20 when results are published and someone other than the publisher ran them or can rerun them, 10 for the publisher's own results, 0 for none.
- 0 to 15, guidance for agent use. 15 for documented prompt-injection guidance or an instruction hierarchy the model was trained to follow, 10 for a statement of what its trained refusals cover and the safeguards to add around it, 5 for general cautions, 0 for none.
- 0 to 15, licence risk. 15 to start, 5 off for each of these, a use policy or field-of-use limits beyond the law, terms the licensor can change or revoke, a revenue or user threshold, and a licence that changed without a dated note or whose link is broken. Not below 0.
- 0 to 10, a security programme. 5 for a way to report a flaw (security.txt, a disclosure policy or a security contact), 5 for advisories handled in public.

When our reference-machine runs start, measured robustness (prompt injection with AgentDojo, over-refusal with XSTest) joins Risk.

## 3. Documentation and upkeep, 40 out of 100, up to 9.2 more on the total

Why it scored 40: Model card 5 of 20, two of the seven items. Training data 5, described in the report with no data published and no cut-off. Last change 20, the card changed on release day. Responsiveness 0, the one question has no reply yet. Version policy 5, a first release. File health 5, the licence metadata is inconsistent with what the card and the Decision Index said.

The checklist (https://www.anchorterminal.com/benchmark/#checklist-upkeep):

- 0 to 20, the model card. Intended use, limitations, architecture, parameters, context length, languages and knowledge cut-off. 20 for all seven, 3 off for each that's missing.
- 0 to 20, training data. 20 for the data published or listed with sources and sizes, 15 for a data card that names the sources or the mix, the cut-off and the filtering, 10 for a general description with a cut-off, 5 without one, 0 for none.
- 0 to 20, the last change to the files, the template or the reference code. 20 within 30 days, 15 within 90, 10 within 180, 5 within a year, 0 older.
- 0 to 20, responsiveness. The publisher's replies to the 40 most recent questions and reports in the repository's discussions or issues. 20 for most, 10 for some, 5 for a few, 0 for none.
- 0 to 10, version policy. 10 for earlier versions kept in their own repositories with a dated note when one is replaced, 5 for earlier versions kept without notes or a first release, 0 when files were replaced in place without a note.
- 0 to 10, file health. 10 when config, tokenizer, template and card agree and no report of a broken file waits without a reply, 5 when one does, 0 when the files contradict the card.

## 4. Agent interface, 45 out of 100, up to 8.5 more on the total

Why it scored 45: Read as a decision model, by what it takes and returns. Chat template 10 of 20, the base's template is there but the decision prompt is built in code. Tool-calling format 10, the request and answer shapes are only in engine.py and OneJev's schema, while Baseten documents them for its hosted API. Trained tool calling 20, several questions per request from one prefill. Structured output 5, claimed on the card without how. Reasoning control 0, it doesn't think. Context 0, trained at 16,384 tokens and refused beyond.

The checklist (https://www.anchorterminal.com/benchmark/#checklist-interface):

What a model and a developer need to talk to it. Speech, image and other models with no chat interface read the tool-calling and structured-output lines for what the model takes and returns (voices, timestamps, streaming, typed outputs), and the note says how.

- 0 to 20, the chat template. 20 for a template in the repository (Jinja, in the tokenizer config or its own file), 10 when the prompt format is only given as code you run.
- 0 to 25, the tool-calling format. 25 when the publisher documents how tools are declared and how calls come back, with the parser to use in a named runtime, 10 when only the template or a runtime's own recipe says it.
- 0 to 20, tool calling the model was trained for. 20 with several calls in one turn, 10 for one call at a time.
- 0 to 15, structured output. 15 when JSON or JSON Schema output is documented for the model with how to ask for it, 5 when the card claims it without saying how.
- 10, reasoning control. Thinking that can be turned off or budgeted per request, with where the reasoning text comes back.
- 0 to 10, context. 10 for 128K tokens or more trained, 5 for 32K to 127K. Context reached only by scaling the position encoding counts one band down.

## Deductions

Each comes off the total. A fixed and documented problem counts for less at the next check.

- 21 October 2026. Baseten's blog says elephant-1 is tied for first with d1-ultra on the Artificial Analysis Decision Index, and its library page gives elephant-1 a score of 65.2 there. Artificial Analysis's own page that day lists d1-ultra at 65.25 and Elephant 9B at 64.69, second of the decision models, and states no tie rule. 65.2 is the Base Labs report's own replica figure. The two only match when rounded to whole numbers. Sources: https://www.baseten.co/blog/the-best-decision-models-in-the-world-mammoth-1-and-elephant-1/ , https://www.baseten.co/library/elephant-1/ and https://artificialanalysis.ai/evaluations/artificial-analysis-decision-index (-2)

## What we couldn't check

What we couldn't read counted as absent. Publishing it on a page a plain HTTP fetch can read (not only in a browser) lets the next check count it.

- Whether `qev serve` after setup.sh serves this checkpoint as trained couldn't be checked without running it. Base Labs documents only the Decision Index pipeline, and that only in the card's earlier version
- Whether Baseten's hosted Elephant 9B, which Artificial Analysis benchmarked, is the same checkpoint as the Hub's files isn't stated
- The deduction for the tied-for-first claim rests on Artificial Analysis's published scores (65.25 against 64.69). If the founder reads a match at whole numbers as a tie, it lapses
- MIT is named only in the metadata, with no licence text or copyright line, and the Apache-2.0 base's licence isn't carried in the repository. Whether that's enough to count as open-source is the founder's call
- The Decision Index excluded one private test for signs of training overlap, against the report's zero-contamination statement. Recorded as a finding, not deducted

## Weaknesses

- Serving it needs OneJev at a pinned commit plus the repository's serve.patch, with no released runtime support
- The licence was proprietary until release day, and MIT now appears only in the metadata with no licence file
- Trained at 16,384 tokens, and longer requests are refused
- The card is two sentences, with no limitations, languages or cut-off
- No quantised build, no safety results and no prompt-injection guidance

## What costs an agent a turn today

The notes we give agents before they call it. Each one is a workaround an agent shouldn't need.

- 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 8000` is 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

## What the review panel asked for

- a licence file
- safety and injection guidance
- a LICENSE file and dated change note
- documented serving steps on the card
- Document the request schema on the card
- Restore the run instructions
- A LICENSE file
- A quantised build

## When it's done

Send what changed and where it's published as a dispute (https://www.anchorterminal.com/builders/#disputes, or `POST https://www.anchorterminal.com/api/v1/contact` with `"kind": "dispute"`). Disputes are answered in public, and the listing is checked again by the same checklist. Paying for an audit or a listing claim changes nothing here.