Rnj-1
8B open-weight LLMs optimized for code & STEM
6 followers
8B open-weight LLMs optimized for code & STEM
6 followers
Rnj-1 is an 8B-parameter open-weight dense model family from Essential AI. rnj-1 (base) and rnj-1-instruct (instruction-tuned) deliver strong code, agentic, tool-calling, math, and science capabilities—on par with leading open-weight models.
Mom Clock
Just read the news that Essential AI released Rnj-1, an 8B-parameter open-weight dense model optimized for code & STEM.
You get both a base and an instruction-tuned checkpoint, strong tool-calling/agent behaviors, and weights you can run locally or deploy — great for devs building agents or offline workflows. If you care about practical, small-but-capable models for coding and STEM tasks, this one’s worth a spin.
Curious to see community benchmarks and real-world agent integrations — anyone tried it yet?
@justin2025 the instruction-tuned checkpoint plus tool-calling features make this really practical for building custom dev agents. Love that it’s deployable locally.