README.md

---
license: apache-2.0
base_model: mistralai/Devstral-Small-2-24B-Instruct-2512
library_name: peft
tags:
- mlx
- lora
- peft
- ailiance
- devstral
- python
language:
- en
- fr
pipeline_tag: text-generation
---

# Ailiance — Devstral-Small-2-24B-Instruct python LoRA

LoRA adapter fine-tuned on `mistralai/Devstral-Small-2-24B-Instruct-2512` for **python** tasks.

> Maintained by **Ailiance** — French AI org publishing EU AI Act aligned LoRA adapters and datasets.

## Quick start (MLX)

```python
from mlx_lm import load, generate

model, tokenizer = load(
    "mistralai/Devstral-Small-2-24B-Instruct-2512",
    adapter_path="Ailiance-fr/devstral-python-lora",
)

print(generate(model, tokenizer, prompt="..."))
```

## Training

| Hyperparameter   | Value                  |
|------------------|------------------------|
| Base model       | `mistralai/Devstral-Small-2-24B-Instruct-2512`     |
| Method           | LoRA via `mlx-lm`      |
| Rank             | 16            |
| Scale            | 2.0           |
| Alpha            | 32           |
| Max seq length   | 2048  |
| Iterations       | 500           |
| Optimizer        | Adam, LR 1e-5          |
| Hardware         | Apple M3 Ultra 512 GB  |

## Training data lineage

Derived from the internal **eu-kiki / mascarade** curation. All upstream samples
are synthetic, permissively-licensed, or generated from Apache-2.0 base resources.
See the [Ailiance-fr catalog](https://huggingface.co/Ailiance-fr) for related cards.

## Benchmark roadmap

This LoRA has **not yet been evaluated** through `electron-bench` (the current
pipeline supports `gemma-4-E4B` base only). Training was completed with the
standard `mlx-lm` LoRA trainer (rank 16, alpha 32, scale 2.0, AdamW
LR 1e-5, 500 iters) — full hyperparameters are in the `Training` table above.

Planned evaluations:

- Perplexity on the validation split of the training data
- Functional benchmark on **devstral**-specific tasks
- Comparison vs base `mistralai/Devstral-Small-2-24B-Instruct-2512`

Track progress: [ailiance-bench issues](https://github.com/ailiance/ailiance-bench/issues).

For reference benchmarks on the `gemma-4-E4B` base, see the
[base-vs-LoRA matrix](https://github.com/ailiance/ailiance-bench/blob/main/bench-results/compare_base_vs_lora.md).

## License chain

| Component                         | License           |
|-----------------------------------|-------------------|
| Base model (`mistralai/Devstral-Small-2-24B-Instruct-2512`)        | apache-2.0    |
| Training data (internal Ailiance curation (synthetic + permissive sources))         | apache-2.0      |
| **LoRA adapter (this repo)**      | **apache-2.0**|

_All upstream components are Apache 2.0 / MIT — LoRA inherits permissive terms._

## EU AI Act compliance

- **Article 53(1)(c)**: training data licenses preserved (per-dataset cards declare upstream licenses).
- **Article 53(1)(d)**: training data summary — see upstream dataset cards on Ailiance-fr.
- **GPAI Code of Practice (July 2025)**: base `mistralai/Devstral-Small-2-24B-Instruct-2512` released under apache-2.0.
- **No web scraping by Ailiance**, **no licensed data**, **no PII**.
- Upstream Stack Exchange content (where applicable) is CC-BY-SA-4.0 and propagates to this adapter.

## License

LoRA weights: **apache-2.0** — see License chain table above for derivation rationale.

## Citation

```bibtex
@misc{ailiance_devstral_python_2026,
  author    = {Ailiance},
  title     = {Ailiance — Devstral-Small-2-24B-Instruct python LoRA},
  year      = {2026},
  publisher = {Hugging Face},
  url       = {https://huggingface.co/Ailiance-fr/devstral-python-lora}
}
```

## Related

See the full [Ailiance-fr LoRA collection](https://huggingface.co/Ailiance-fr).


## Bench comparison (2026-05-11)

### Base model (Devstral-Small-2-24B-MLX-4bit) capability

| Task | Score | Notes |
|---|---:|---|
| GSM8K-CoT flex EM | **0.96** | W3 lm-eval-harness (--limit 100) |
| ARC-Easy acc / acc_norm | **0.80 / 0.75** | |
| MMLU-Pro Computer Science | **0.64** | |

Source: <https://github.com/ailiance/ailiance/tree/main/output/lm-eval-base-2026-05-11>

### This LoRA (tuned) — bench PENDING

Will include kicad-sch / iact-bench validators + W3 lm-eval delta. See spec for
methodology:
<https://github.com/ailiance/ailiance-bench/blob/main/docs/superpowers/specs/2026-05-11-kicad-sch-gap-design.md>

## Upstream base model — official evaluations

This LoRA fine-tunes [`mistralai/Devstral-Small-2-24B-Instruct-2512`](https://huggingface.co/mistralai/Devstral-Small-2-24B-Instruct-2512),
Mistral's coding-specialist LLM. Headline software-engineering benchmarks
from the upstream model card:

| Benchmark                | Devstral Small 2 (24B) | Devstral 2 (123B) | DeepSeek v3.2 (671B) | Claude Sonnet 4.5 |
|--------------------------|-----------------------:|------------------:|---------------------:|------------------:|
| **SWE Bench Verified**   | **68.0 %**             | 72.2 %            | 73.1 %               | 77.2 %            |
| **SWE Bench Multilingual** | **55.7 %**           | 61.3 %            | 70.2 %               | 68.0 %            |
| **Terminal Bench 2**     | **22.5 %**             | 32.6 %            | 46.4 %               | 42.8 %            |

(For reference, GPT-5.1 Codex High: 73.7 % SWE Verified · 52.8 % Terminal Bench 2.)

Devstral Small 2 (24B) is competitive with much larger open models on
SWE Bench Verified (e.g. matches GLM-4.6 at 355B). Architecture uses
rope-scaling per Llama 4 + Scalable-Softmax ([arXiv:2501.19399](https://arxiv.org/abs/2501.19399)).

**Source:** [official Devstral-Small-2-24B-Instruct-2512 model card](https://huggingface.co/mistralai/Devstral-Small-2-24B-Instruct-2512).

> **Reading these alongside this LoRA:** Devstral Small 2 is a strong
> coding base. This LoRA inherits its SWE-Bench performance and adds
> language- or domain-specific specialization.