Function calling performance varies by provider, even for the same model. See how inference providers Fireworks, Together AI & SambaNova performed.