Inkling Small
- Source
- thinkingmachines.ai
- Author
- Thinking Machines
- Date
Context
Thinking Machines presents Inkling-Small as a smaller alternative to its earlier Inkling model, with a different balance of capabilities. It has 276 billion parameters in total (12 billion active per step) against Inkling's 975 billion (41 billion active). The company reports stronger reasoning and agentic coding results for the smaller model, while the larger one retains an advantage in factuality and knowledge coverage. Release and comparison.
The smaller model also received a revised training recipe, including learning from Inkling and further training on coding tasks. Its results therefore compare differently trained models, rather than showing the effect of reducing size alone. Developers receive and access to through Tinker, allowing them to evaluate that tradeoff for their own tasks. Training and availability.
- open weights — A model whose trained parameters are published for anyone to download and run — unlike API-only models you can access but never possess.
- fine-tuning — Taking a trained model and training it a bit more on your own examples so it gets better at one specific job.
- context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
Checking sign-in…
Loading comments…