Over the last decades, climate science has branched out into many smaller expert communities across the carbon cycle, radiative forcings, climate feedbacks or ocean heat uptake domains. Our best tools to capture state-of-the-art knowledge are the increasingly complex fully coupled Earth System Models (ESMs). However, computational limitations and the structural rigidity of ESMs mean that the full range of uncertainties are difficult to capture with multi-model ESM ensembles or single ESM perturbed parameter ensembles. The tools of choice are hence more computationally efficient reduced complexity models (RCMs), which are structurally flexible and can span the response dynamics across a range of domain-specific models and/or ESM experiments. Here, we provide the first comprehensive intercomparison of multiple RCMs that are probabilistically calibrated to key benchmark ranges from specialised research communities. This exercise constitutes Phase 2 of the Reduced Complexity Model Intercomparison Project (RCMIP Phase 2). We find that even if RCMs perform similarly against historical benchmarks, their future projections can still diverge. Under the low-emissions SSP1-1.9 scenario, across the RCMs, median 2081-2100 warming projections range from 1.1 to 1.4{degree sign}C while median peak warming projections range from 1.3 to 1.7{degree sign}C (relative to 1850-1900, using an observationally-based historical warming estimate of 0.8{degree sign}C between 1850-1900 and 1995-2014). Our findings suggest that users of RCMs should carefully evaluate the RCM they are using, specifically its skill against key benchmarks and consider the need to include future projections benchmarks either from ESM results or other assessments to reduce such divergence.