Kimi K3 vs OTel-2.0-LLM-31B-IT

Kimi K3 41

The model card earns 11, covering intended use, architecture, parameters and context, but not limitations, languages or a knowledge cut-off. Training data earns 5 for a general description with no cut-off. The last change earns 15, an encoding_k3.py update on 20 August, 74 days before 2 November and within 90; the GitHub repository's last commit is 6 August. Responsiveness earns nothing, with no Moonshot reply in the 40 most recent discussions. Version policy earns 5 for a first release. File health earns 5, because a report that the tokeniser goes past the declared vocabulary waits without a reply.

OTel-2.0-LLM-31B-IT 74

Model card 14, Intended use, a long limitations section, the architecture and training method, the parameters (31B) and the language (English) are on the card. Training data 15, the card names the sources (3GPP, ETSI, ITU, GSMA, CAMARA, O-RAN and TM Forum material provided by GSMA), the overall sizes (15 billion raw tokens, over 1 trillion processed, about 440 billion used), the mix of example types and the processing tool, and lists what was left out. Last change 20, the last commit, on 1 November 2026, updated the card's links, resources, citation and contacts, nine days before this check. Responsiveness 20, Of the 3 discussions that aren't pull requests (7 August to 28 October 2026), 2 have a reply from the repository's owner, which is most. Version policy 0, the weights were replaced in place: uploads on 27 July, 3 August and 1 September 2026 under one name, with a standing notice on the card that they may change and that users should pin a revision, and no dated note of what changed. File health 5, the card describes 15 shards of about 62.6 GB while the repository still holds an earlier 14-shard set beside them, 126.8 GB in all. 14 + 15 + 20 + 20 + 0 + 5 = 74.

Kimi K3 65

File safety earns 15, safetensors only with no pickle files, less 10 because the documented loading path runs Moonshot's Python with trust_remote_code and builds prompts with encoding_k3.py. Provenance earns the full 15, files in the organisation Moonshot links to with a SHA-256 for every shard. Safety evaluation earns 20 for the cyber assessment the UK AI Security Institute and CAISI ran and published, which found its safeguards didn't stop exploit development. Agent guidance earns 5 for the blog's general cautions about proactiveness and passing back thinking. Licence risk earns 10, losing 5 for the model-as-a-service revenue threshold and nothing else. The security programme earns nothing, with no security.txt, disclosure policy, SECURITY.md or public advisories found.

OTel-2.0-LLM-31B-IT 55

File safety 25, safetensors only, with no .bin, .pt or .pth files and no Python files. Provenance 5, the files sit in a personal account, farbodtavakkoli, not in an organisation AT&T's or GSMA's site links to: GSMA's article names AT&T as the releaser without linking the repository, and the GitHub repository that links it is the same person's (0). Safety evaluation 0, no safety results on the card. Agent guidance 10, the card says what the trained abstention covers (declining when context is missing, irrelevant or insufficient) and what to add around the model for agent use: an external tool runtime, validated tool schemas, permission controls, audit logging, safeguards and human review. Licence risk 15, Apache 2.0 has no use policy, field-of-use limit, threshold or change clause, nothing is added on the card, and the licence link resolves (200 on 10 November 2026). Security programme 0, no way to report a flaw found. 25 + 5 + 0 + 10 + 15 + 0 = 55.

Kimi K3 65

The chat template earns 10, because there is no Jinja template and the prompt format exists only as encoding_k3.py, code you run. The tool-calling format earns 10, since Moonshot documents the tools channel but the parser name comes only from vLLM's and SGLang's own recipes. Trained tool calling earns 20 for indexed parallel calls in one turn. Structured output earns 5, as the card names it among its linked guides without saying how to ask the weights for it, and the how sits in hosted-API docs. Reasoning control earns 10, with per-request effort documented on the card and reasoning returned as reasoning_content. Context earns 10 for 1,048,576 tokens trained in.

OTel-2.0-LLM-31B-IT 60

Chat template 20, chat_template.jinja in the repository, 19,300 characters, the publisher's own. Tool-calling format 10, only the template says how tools are declared and how calls come back. Trained tool calling 20, the template loops over tool_calls, so it renders several calls in one turn, and the card says the training mix included general-purpose agentic tool-calling examples, while none were specific to telecoms. Structured output 0, the card doesn't document JSON or JSON Schema output. Reasoning control 0, the card says the model wasn't trained to produce or consume Gemma's thought channel and answers directly, so there is no thinking to switch or budget. Context 10, 262,144 positions in the text config, the base's native context. 20 + 10 + 20 + 0 + 0 + 10 = 60.