Alpha
Contact
Support · Luciole 1.1 Instruct

Will it last?

Ownership levelPartialnone·limited·partial·substantial·fullAnalytical input D ยท 51.2/100

This page is a projection of the one entry record, the Transparency factor that Support 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

Common problems & fixes

  • Ollama output is poor: use the official GGUF files and the published Modelfile (ev-card-instruct).
  • Tool calls are not returned in vLLM: start the server with --enable-auto-tool-choice --tool-call-parser hermes.
  • The Thinking models need --reasoning-parser qwen3 and a recent vLLM (ev-card-thinking).
  • A user has opened a discussion on the Instruct repo about an extra vLLM parser configuration (ev-hf-platform).
  • The Modelfile comment names "Luciole-1B-Instruct-1.2". The file loads the 1.1 GGUF it points to, so ignore the comment.
  • The 8B fails to load without repository code: the hybrid architecture ships modeling_nemotron_h.py in the repo.

Versions, changelog & cadence

The base models appeared on 2026-02-17 and Instruct 1.1 on 2026-07-07, with GGUF and FP8 builds on 2026-07-08. Thinking-1.1 followed on 2026-09-09, Audio-1.0 on 2026-09-18, KD-Math on 2026-09-28 and 2026-10-13, and FP8 builds of the Thinking models on 2026-11-05 (ev-hf-platform).

The Instruct repo has nine commits since the initial release, the latest on 2026-11-12 for transformers 5 compatibility. There is no changelog file. The card calls Instruct-1.1 the first iteration and says later iterations with more French data are planned. The cards give no schedule.

Security / vulnerability disclosure

The SECURITY.md of the Luciole-Training repository asks for private reports to contact@openllm-france.fr. It covers the code, and says issues in a published model or dataset can be raised on Hugging Face. It sets no response time and offers no signed disclosure process (ev-code-repo).

Community & support channels

The Hugging Face repo has one open discussion. The consortium site lists contact@openllm-france.fr for contributions and a data-deletion form for the datasets (ev-hf-platform, ev-website). No commercial support channel is listed.

Tracked known issues

The publisher lists these (ev-card-instruct, ev-card-thinking):

  • hallucination, and limited memorisation at 23B;
  • post-training almost entirely in English;
  • 16,384-token post-training sequences;
  • tool use in progress;
  • for the Thinking model, no identity training and over-refusal of tool calls.

No public issue tracker for the models was found. The training-code repository is GPL-3.0 and has a public issue tab that was not reviewed.

How this scores

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

2

TransparencyDo you know what it is: weights, training, behaviour, and legible terms?

Strong

Weights, training and evaluation code, the pretraining corpus and the post-training data are public, and the cards are detailed. Gaps are the unpublished technical report and the external server for three web subsets (ev-pretrain-data, ev-posttrain-data, ev-code-repo).

How this scores (AOI sub-dimensions)
Provenance2/5how well we can trace and verify what went into the modelSafetensors only, a clean Hugging Face scan, LFS hashes and one canonical source give about 4 of 8 checklist boxes.
Governance3/5how accountable and well-documented the publisher isLINAGORA SAS is a named company, the project has Bpifrance funding and GENCI compute, and releases came in February, July, September, October and November 2026.
What this means for adoptionYou own the use of the weights. Apache-2.0 is declared, the repos are ungated and your data stays on your hardware. Training code and both datasets are public, so most of what the model is can be inspected. You cannot rely on its safety, because no safety evaluation or guard model exists. The level rises when a safety evaluation, a LICENSE file in the Instruct repos and a technical report are published.

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.

Model cardread2026-11-19
The 23B Instruct card declares Apache-2.0 and names LINAGORA SAS as copyright holder.
Model cardread2026-11-19
The Thinking-1.1 card opens with a note that the model 'has undergone safety alignment during the DPO thinking phase' and can still produce unsafe output.
Model cardread2026-11-19
The Luciole-23B-Base card declares Apache-2.0 and lists French, English, Italian, German, Spanish, Portuguese, Dutch and Arabic with token shares.
Licenceread2026-11-19
The Base repos contain a LICENSE.md holding the full Apache License 2.0 text with the template front matter and an unfilled 'Copyright [yyyy] [name of copyright owner]' line.
Licenceread2026-11-19
The NOTICE file of the Luciole-Training repository says the Luciole models, including Instruct-1.1 in all three sizes, are the property of LINAGORA SAS and released under Apache-2.0.
Licenceread2026-11-19
The LICENSE file of the Luciole-Training repository is the standard GNU General Public License version 3 text.
Documentationread2026-11-19
The Luciole-Training repository has 532 commits and holds data processing, tokenizer training, NeMo pretraining, conversion, NeMo and NeMo-RL fine-tuning folders and a lighteval-based evaluation folder.
Documentationread2026-11-19
The OpenLLM-France organisation is not verified on Hugging Face, has 20 members and lists 38 model repos.
Documentationread2026-11-19
Luciole-PostTraining-Dataset-1.1 is public and ungated under CC-BY-SA-4.0.
Documentationread2026-11-19
Luciole-Training-Dataset is public and ungated under CC-BY-SA-4.0, with per-source licences listed.
Model cardread2026-11-19
The 8B and 1B Instruct cards match the 23B card.
Technical_reportunverified2026-11-19
No Luciole technical report exists.
Model cardread2026-11-19
The benchmark figure on the 23B Instruct repo, made by the publisher, compares Luciole with Gaperon-1125-24B-SFT, EuroLLM-22B-Instruct-2512 and two Mistral-Small-24B models.
Vendor announcementread2026-11-19
The consortium site calls the project 'totalement open source', says it is funded by Bpifrance for two years with nine partners, and says it republishes training datasets.
Documentationread2026-11-19
The Nemotron-Instruction-Following-Chat-v1 card says its data came from GPT-OSS-120B and Qwen3-235B and is CC-BY-4.0.
Documentationread2026-11-19
RedTeachingBench is a French false-premise benchmark with 1,000 prompts for teachers using LLMs, under CC-BY-SA-4.0.