Fusion API achieves Fable-level intelligence at half the price, OpenRouter claims
What's the deal? On 13 June 2026, OpenRouter introduced its Fusion API, which it bills as "the smartest compound model in the market". Rather than routing a request to a single model, Fusion combines the responses of several large language models into one "fusion panel" answer. OpenRouter says the approach reaches the intelligence level of Anthropic's Claude Fable 5 at roughly half the price.
The benchmark. In the company's published results, fusion panels consistently outscored the same models run solo: a Fable 5 + GPT-5.5 panel topped the chart at around 69%, ahead of an Opus 4.8 + GPT-5.5 + Gemini 3.1 Pro panel (~68%) and an Opus 4.8 + GPT-5.5 panel (~67.5%), while individual models such as GPT-5.5 and Claude Opus 4.8 solo scored in the high 50s to low 60s.
Why it matters now. The positioning has gained relevance since Anthropic pulled Claude Fable 5 from availability. With the model that anchored the benchmark no longer directly accessible, a fusion panel that approximates Fable-level intelligence becomes more useful, not less — offering a route to comparable performance for buyers who can no longer reach the original model, even if the exact Fable 5 + GPT-5.5 panel that topped the chart is no longer reproducible as published.
The bigger picture. The launch extends OpenRouter's positioning as a unified inference layer across hundreds of models. The company recently raised a $113M Series B at a $1.3B valuation.
Read more: OpenRouter on X
