OpenRouter has introduced Fusion, an API that sends a query to several models, compares the answers via a judge model, and synthesizes the best answer. OpenRouter used the DRACO framework for benchmarking, a deep research evaluation suite by Perplexity AI involving 100 research assignments in ten areas, each evaluated on approximately 39 weighted factors, with negative marks for wrong answers. The fusion of Fable 5 and GPT-5.5 into one panel, using Claude Opus 4.8, yielded 69.0 percent in DRACO, outperforming any individual model tested. Fable 5 achieved 65.3 percent on its own. A budget panel fusing Gemini 3 Flash, Kimi K2.6, and DeepSeek V4 Pro gave 64.7 percent, outperforming GPT-5.5 at 60.0 percent and Claude Opus 4.8 at 58.8 percent. A single model, Claude Opus 4.8, fused with itself as a two-model panel scored 65.5 percent, compared to 58.8 percent for Opus running alone, a 6.7-point jump. OpenRouter found that panels with web search inadvertently located the DRACO grading rubric and fixed this by blocking the domains.