On HealthBench, GPT-6 Sol (max) writes the shortest answers (~1,000 characters) and has the highest raw Communication Quality score (81%), but the lowest Completeness score (34%). Context Seeking or Global Health is the lowest-scoring theme for 13 of the 14 models
On HealthBench, GPT-6 Sol (max) writes the shortest answers (~1,000 characters) and has the highest raw Communication Quality score (81%), but the lowest Completeness score (34%). Context Seeking or Global Health is the lowest-scoring theme for 13 of the 14 models
Original publication and rights belong to the source.
