Are there benchmarks of this / what’s the best way to compare it against paid models? With all the rate limiting in Claude/Copilot/etc, running locally is more

Are there benchmarks of this / what’s the best way to compare it against paid models? With all the rate limiting in Claude/Copilot/etc, running locally is more and more appealing.