Alpha
Contact
Use · MiniCPM5-1B

Is it good enough?

Ownership levelPartialnone·limited·partial·substantial·fullAnalytical input C ยท 56.8/100

This page is a projection of the one entry record, the Reliability factor that Use covers. The full verdict is set by all four factors together, floor-weighted so the weakest caps the whole.

Which domain expands which factor
  • AssessUse & modify + Transparency
  • ImplementData control + Reliability
  • UseReliability
  • SupportTransparency

Capabilities & modalities

Text in, text out (ev-aa, ev-card). One checkpoint serves a fast assistant mode and a deliberate reasoning mode, switched by enable_thinking. The publisher highlights code, maths, instruction following and tool use. Its figures (Thinking mode, unverified): MATH-500 91.60, LCB-v6 33.52, AIME-2026 40.42, IFEval 80.41, MMLU-Pro 48.85, tau2-Bench Telecom 79.53 (ev-bench). Non-thinking performance is not published. Treat these as claims until reproduced.

Context window & long-context behaviour

131,072 tokens in config.json and the GGUF metadata, with rope_theta 5,000,000 and no rope scaling (ev-repo-files). The card says "native long-context support" but publishes no long-context result for the 1B (the series radar chart with a long-context axis belongs to the 2B card). For a 1B model, test retrieval and instruction following at the real context lengths. Cap the context on small devices, because the KV cache grows to about 3.2 GB at the full window.

Prompt format & chat template

The tokenizer ships chat_template.jinja, a ChatML-style template (<|im_start|>role ... <|im_end|>). Use apply_chat_template with enable_thinking=True (the template appends <think>) or enable_thinking=False (it appends an empty think block). Tool definitions are injected in <tools></tools> and calls use <function name="..."><param name="...">...</param></function>, with tool results wrapped in <tool_response> (ev-repo-files). Sampling per mode is on the card: 0.9 or 0.7 temperature, top_p 0.95. The generation config defaults to temperature 0.9. EOS ids are 1 and 130073.

Language coverage

English and Chinese. The card metadata lists en and zh, a Chinese README is provided, and the training datasets are English and Chinese (ev-card, ev-datasets). No other language is documented or evaluated by the publisher, and the benchmark table has no multilingual row. Community fine-tunes for other languages exist on the Hub but are not evidence of base-model capability. Any third language is unverified and needs testing.

Function / tool calling

Native. The model emits XML-style function calls; SGLang's minicpm5 parser converts them to OpenAI-style tool_calls, and the repository includes tool_parsers/minicpm5xml_tool_parser.py (ev-card, ev-github). On the publisher table BFCLv4 is 25.15, level with Qwen3-0.6B (25.43) and Qwen3.5-0.8B (25.53) and above LFM2.5-1.2B (10.60), while tau2-Bench Telecom is 79.53 (ev-bench). The card's claim of a tool-use lead rests mainly on the second result. Validate every call before executing it.

How this scores

The ownership factor this domain covers, drawn from the one entry record.

3

ReliabilityIs it reliable and good enough for the job?

Weak

Operational support is broad (4), but there is no safety evaluation, policy or guard model (safety 1), so the model is unsafe to deploy as presented until the deployer supplies that layer.

How this scores (AOI sub-dimensions)
Operational4/5how practical it is to run, serve and maintain in productionStandard LlamaForCausalLM with no custom code, listed support for Transformers, vLLM, SGLang, llama.cpp, Ollama (local Modelfile), LM Studio, MLX, LiteRT-LM and ArcLight, official GGUF and MLX conversions, cookbooks with file sizes, and a 131,072-token context in config.
Safety1/5whether misuse risks are evaluated and guardrails are providedThe only safety text is a three-sentence limitations paragraph saying the model may produce inaccurate, biased or unsafe output and that users must apply safeguards.
What this means for adoptionUse of these weights is fully owned: Apache-2.0, ungated, runnable on a phone or laptop, with Base and SFT checkpoints and public pretraining data to adapt from. Ownership is partial because the reliability factor is weak: OpenBMB publishes no safety evaluation, policy or guard model. The SFT data card is gated, and the only benchmark is the publisher's own table. The model suits local assistants and fine-tuning behind the deployer's own guardrails and safety tests. Ownership moves to substantial once a safety evaluation exists.

Sources

The same evidence records as the entry sheet. Read means the text was verified; unverified means it is known to exist but not yet read.

Licenceread2026-11-19
The LICENSE file in the OpenBMB/MiniCPM repository, which the MiniCPM5-1B model card links as the licence for "this repository and MiniCPM model weights", is the standard Apache License 2.0 text.
Licenceread2026-11-19
Hub card metadata for openbmb/MiniCPM5-1B, -SFT, -Base, -GGUF, -MLX and every MiniCPM5-2B checkpoint reads license apache-2.0 with gated false.
Licenceread2026-11-19
The legacy README for earlier MiniCPM generations says model weights "must strictly follow" a custom General Model License (GML), with commercial use needing authorisation from cpm@modelbest.cn by registration.
Model cardread2026-11-19
The MiniCPM5-1B model card describes a dense LlamaForCausalLM with 1,080,632,832 parameters, 24 layers, GQA with 16 query and 2 key-value heads, and a context length of 131,072.
Documentationread2026-11-19
The Hub repository holds one safetensors shard (2,161,290,912 bytes, LFS sha256 7ab8fd86...), no pickle and no LICENSE file.
Documentationread2026-11-19
The dataset cards for Ultra-FineWeb (about 1T English and 120B Chinese tokens), Ultra-FineWeb-L3 and UltraData-Math show that all three are public, ungated and Apache-2.0 tagged.
Documentationunverified2026-11-19
UltraData-SFT-2605 exists on the Hub (API metadata: gated auto, apache-2.0, 1,510 files, created 2026-05-21) but its card returns 401 "restricted ...
Technical_reportread2026-11-19
The card cites two arXiv papers.
Vendor announcementread2026-11-19
The publisher benchmark image on the model card (Thinking mode, three small baselines) reports an average of 42.57 vs Qwen3-0.6B 26.77, Qwen3.5-0.8B 25.14 and LFM2.5-1.2B 35.61.
Third-party analysisread2026-11-19
Artificial Analysis lists MiniCPM5-1B as an open-weights OpenBMB model, Apache 2.0, text only, 128k context, released 25 May 2026, with no publicly available API and no Intelligence Index score or other benchmark result on its model page.
Documentationread2026-11-19
The OpenBMB/MiniCPM repository holds a release log (MiniCPM5-1B 2026-05-19) and names the institutions (Modelbest, THUNLP, Renmin University GSAI).
Model cardread2026-11-19
The MiniCPM5-2B card (second model in the series, released 2026-09-07, Apache-2.0, ungated) lists a different and larger dataset set (adds UltraData-Code, UltraData-SFT-Agent-2609, UltraData-RL-2609, UltraX-Preview), so it is a separate release and is not scored in this entry.