MiniCPM-o 4.5

# Fix list: MiniCPM-o 4.5

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/minicpm-o-4-5, the October 2026 research run, assessed 21 October 2026. Grade C, 61 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 MiniCPM-o 4.5: 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. Risk, 40 out of 100, up to 18.5 more on the total

Why it scored 40: File safety 5 of 25, two pickle files sit in the speech stage and the documented load runs repository code. Provenance 15, the files are in the openbmb organisation OpenBMB's GitHub links, with a checksum on every file. Safety evaluation 0, none published. Agent guidance 5, general cautions only, nothing about voice cloning misuse, prompt injection or refusals. Licence risk 15, plain Apache 2.0. Security programme 0, no security.txt, policy or 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.

## 2. Agent interface, 35 out of 100, up to 10 more on the total

Why it scored 35: Chat template 20, a Jinja template in the tokenizer config. Tool-calling format 10 of 25, only the inherited Qwen3 template says how tools work. Trained tool calling 0, OpenBMB never mentions tool use. Structured output 0, not documented. Reasoning control 0, thinking turns on and off per request but nobody says where the reasoning comes back. Context 5, 40,960 tokens. Read as an omni model, it takes images, video and audio and returns text and streaming speech, which the card documents with code.

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.

## 3. Documentation and upkeep, 66 out of 100, up to 5.2 more on the total

Why it scored 66: Model card 11 of 20, limitations, context length and a knowledge cut-off are missing. Training data 5, described by kind without sources, sizes or a cut-off. Last change 20, the card changed 5 days ago. Responsiveness 10, OpenBMB answered half of the 12 questions. Version policy 10, earlier models stay in their own repositories and releases are dated. File health 10, the files agree and the two bug reports have replies.

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. Reach, 90 out of 100, up to 3.8 more on the total

Why it scored 90: Memory 30, OpenBMB's 4-bit GGUF with its vision file is 5.9 GB, about 7 GB with a fifth added. Runtimes 15 of 25, released vLLM, llama.cpp and Ollama run it, while Transformers needs OpenBMB's remote code and SGLang supports only the older 2.6. Builds 15, GGUF at several precisions plus AWQ and GPTQ. Terms 15, plain Apache 2.0. Download 10, nothing gated. Hosted 5, OpenBMB's own 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.

## 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.

- OpenBMB's API at api.modelbest.cn serves this model and isn't listed with us, so there's no sameModel to set; whether to list it is the founder's call
- Ollama is credited on released Ollama's pinned llama.cpp having the 4.5 projector and OpenBMB's namespace build, although OpenBMB's Ollama guide still tells readers to build its fork
- unchecked, GitHub issues on OpenBMB/MiniCPM-o weren't counted, since GitHub's API isn't reachable from this session

## Weaknesses

- Speech output and full duplex need trust_remote_code with Transformers 4.51.0 or OpenBMB's llama.cpp-omni
- Two pickle files (flow.pt and hift.pt) in the speech stage
- No safety results or guidance, though it clones voices from a short clip
- Tool calling, structured output and where the reasoning comes back aren't documented
- 40,960 tokens of context, and OpenBMB answered only half the questions in the discussions

## 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.

- Quickest start for text and images is `ollama run openbmb/minicpm-o4.5`, about 6 GB
- With vLLM 0.16.0 or later pass `--trust-remote-code`, and send `stop_token_ids` [151643, 151645] as OpenBMB's guide does
- vLLM, llama.cpp and Ollama return text only; for speech out use the card's Transformers 4.51.0 code or OpenBMB's llama.cpp-omni
- In llama.cpp pass both the language GGUF and MiniCPM-o-4_5-vision-F16.gguf with `--mmproj`, and use `--jinja --reasoning-budget 0` to turn thinking off
- Keep prompts within 40,960 tokens
- Don't rely on tool calls or JSON output, neither is documented; validate any you get

## What the review panel asked for

- merge Transformers 5 support
- published API prices
- state whether tool calls are supported
- convert flow.pt and hift.pt to safetensors
- a security contact

## 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.