Olala mini 7A1B
AvailableOVHai LLM's first model, launched Nov 19 2026 at the OVHcloud AI Summit and released with open weights under Apache 2.0 at Hugging Face OVHaiLLM/Olala-v1.0-mini-7A1B. A text-only post-trained reasoning model for chat, instruction following and tool use, with four reasoning-effort levels. 6.8B total parameters and about 1B active per token; 36 layers (29 Mamba-3 and 7 attention), each followed by a mixture-of-experts block with 256 routed experts (6 active) and 1 shared expert; 32,688-token context (input and output together). Trained on 12.8T tokens of public datasets and synthetic data on the JUPITER, Jean Zay and MeluXina supercomputers. Pre-training data goes up to January 2026, post-training data up to April 2026. Served on OVHcloud AI Endpoints (model Olala-v1.0-mini-7A1B) and on Blade. Announced Nov 17 2026.
Specifications
- License
- Open source · Apache-2.0
- Weights
- Downloadable
- Architecture
- hybrid
- Parameters
- 6.8B · 1B active
- Context window
- 33K tokens
- Max output
- —
- Knowledge cutoff
- —
- Price (in / out, $/M)
- —
- Modalities
- Text
Benchmarks
No benchmark scores recorded yet. Spotted some? Submit a correction.
Vendor-reported figures are claims until independently verified. See methodology.