PoolsideLaguna
Laguna M.1 base
This 225-billion-parameter base checkpoint precedes instruction tuning and reinforcement learning. It has 23 billion active parameters and an Apache 2.0 license.
What the release provides
It is a text-completion model. The release has no chat template, tool-calling behavior or instruction-following post-training. Use the post-trained M.1 model for a ready-made coding assistant.Poolside: Laguna M.1 base model card
How to use it well
Practical guidance. Use a base checkpoint when the work includes further training. Do not evaluate it as a conversational assistant without accounting for its different training stage.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| Model identifiers | poolside/Laguna-M.1-base |
| Context window | 262,144 tokens Published model context window Poolside: Laguna M.1 base model card |
Related models
- Laguna M.1
Poolside · Laguna
A large open-weight Laguna model for multi-step coding work.
Launch date not confirmedDownloadable · Provider product