701 3 weeks ago

3B model that shouldn't be this good - crushes benchmarks through deep chain-of-thought reasoning