Aleph Alpha open-sources Kolibri-1: 78B German/English reasoning MoE with 1M context and native tool calling
Aleph Alpha released Kolibri-1: a ~78B-total / 3.46B-active MoE reasoning model with native German/English training, up to 1,048,576-token context, and native tool calling, under an open license.
Source: AI 日报（ai6s.net）
What it is
On October 3, European AI company Aleph Alpha released open-source reasoning model Kolibri-1: a sparse MoE with ~78B total / 3.46B active parameters, native German/English bilingual training (with substantially higher German share than mainstream English-dominated models), up to 1,048,576-token context, and native tool calling.
Key design
- Only 3.46B active: substantially cheaper inference than dense 70B-class models
- Native DE/EN: targets European enterprise, especially German manufacturing, legal and public sector
- 1M context: enters the million-token club
- Tool calling protocol: compatible with mainstream function-calling conventions
Comparison
| Dimension | Kolibri-1 | Qwen3.5-35B-A3B | DeepSeek-V4.1-Flash |
|---|---|---|---|
| Total / Active | 78B / 3.46B | 35B / 3B | 67B / 3.5B |
| Context | 1M | 256K | 1M |
| Bilingual focus | DE/EN | ZH/EN | ZH/EN |
| Tool calling | Native | Native | Native |
Local deployment angle
3.46B active means Q4_K_M quantization runs on a single 24 GB card; 1M context significantly inflates KV-cache — 32 GB+ VRAM or CPU offload is recommended. Watch community quantized builds in our data section.