Model reference · open weights
Qwen3-SW-Swap is an open-weight language model from lightonai, listed in the AxForge catalogue. AxForge can bring it up on EU-owned hardware for you on request — with the licence handled where one is required.
About
Qwen3-8B-SW-Swap Qwen3-8B-SW-Swap is a Layer Swap model built on top of lightonai/Qwen3-8B-SW: the middle transformer layers (L13–L22) of the English specialist lightonai/Qwen3-8B-EN have been transplanted into the Swahili native specialist. The resulting model reasons natively in Swahili while inheriting the stronger reasoning core of the English specialist. It is released alongside the paper Rethinking the Multilingual Reasoning Gap with Layer Swap. Model details - Base model: Qwen/Qwen3-8B-Base - Construction: Training-free Layer Swap — layers L13–L22 of Qwen3-8B-EN transplanted into Qwen3-8B-SW - Language: Swahili (CoT and answer) - Context length: 32,768 tokens - Dataset (underlying specialists): lightonai/Dolci-Think-SFT-32B-Multilingual [!NOTE] The model was trained on data derived from allenai/Dolci-Think-SFT-32B, released under the ODC-BY-1.0 license. Related models This model is part of a Swahili specialist trio designed to study the native reasoning gap: Evaluation All scores are mean accuracy (%) on the Swahili version of each benchmark, with sample standard deviation across runs. AIME 24/25 is averaged over 30 runs; the others over 10 runs, using the recommended generation parameters. Benchmarks used: - lightonai/gpqadiamondmultilingual - lightonai/aime24multilingual - lightonai/aime25multilingual - lightonai/HumanEvalPlusmultilingual - lightonai/mgsm-rev2 - CohereLabs/Global-MMLU-Lite Usage Recommended sampling: temperature=1.0, topp=0.95, topk=20, minp=0. Citation If you find our work helpful, feel free to give us a cite.
Summarised from the published model card. Read the full card on the HuggingFace links below.
Specifications
| Maker | lightonai |
|---|---|
| Type | Language models |
| Parameters (lead) | 8.2B |
| Context | 40k tokens |
| Variants | 1 |
| Runs with | transformers |
| Based on | Qwen/Qwen3-8B-Base |
| Released | 2026-05-26 |
| Popularity | 12 downloads / month |
| Licence | Open weights |
How it works
Variants
Open weights ship in several sizes and precisions. One page, all the variants — pick the one that fits your GPU. VRAM figures are estimates from model size.
| Variant | Params | Precision | VRAM | Fits 16 GB | Weights |
|---|---|---|---|---|---|
| Qwen3-8B-SW-Swap | 8.2B | BF16 | ~18.8 GB | ✓ | Weights ↗ |
Using it via the API
Once AxForge deploys qwen3-sw-swap for you, it answers on the OpenAI-compatible API — the same base URL and keys as every other model. (qwen3-sw-swap below is illustrative; you get the exact model name on deployment.)
$ curl -sS https://api.axforge.ai/v1/chat/completions \
-H "Authorization: Bearer $AXFORGE_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"qwen3-sw-swap","messages":[{"role":"user","content":"Hello"}]}'
Licence
Open weights under apache-2.0 — commercial use is permitted. Deploy it on AxForge EU hardware on request. Read the licence ↗