OpenRouter launched Fusion on June 12, a server-side API that distributes prompts to multiple AI models in parallel, then uses a judge model and synthesizer to merge responses into a unified answer. The company claims the system can match Claude Fable 5's performance at roughly half the cost, based on testing using Perplexity's DRACO benchmark where a budget panel of models scored 64.7% compared to Fable 5's 65.3%. The launch came shortly after Anthropic suspended Fable 5 and Mythos 5 last week following a U.S. export control directive citing a disputed jailbreak finding, with OpenRouter positioning Fusion as an alternative offering "Fable-level intelligence at half the price."
OpenRouter Fusion Processes Prompts Through Multi-Model Panel Architecture
When a user sends a prompt to Fusion, OpenRouter distributes it to a panel of models in parallel, with each model receiving web search and bash tools. A judge model then extracts consensus points, contradictions, and blind spots from every response. After this analysis phase, a synthesizer—Claude Opus 4.8 by default—writes the final answer grounded in that analysis. The entire process occurs server-side. Users can swap their model string to "openrouter/fusion" for a default panel, add a fusion tool so their own model calls it selectively, or build a custom panel in the Fusion chatroom with no code.
Budget AI Panel Scores 64.7% on DRACO Benchmark Against Fable 5's 65.3%
OpenRouter tested Fusion on DRACO, Perplexity's benchmark built from real user deep research requests. Fable 5 paired with OpenAI's GPT-5.5 and synthesized by Opus topped the chart at 69%. Solo Fable scored 65.3%, though seven of its 100 tasks never ran because its own content filters blocked them. The budget combination of Gemini 3 Flash combined with open-source Chinese models Kimi K2.6 and DeepSeek V4 Pro, fused and synthesized by Opus, hit 64.7%—beating solo GPT-5.5 (60%) and solo Opus 4.8 (58.8%) and landing within one percentage point of Fable at roughly half the cost. Pairing Opus 4.8 with a separate instance of itself scored 65.5%, a 6.7-point jump over solo Opus. OpenRouter states roughly three quarters of that improvement comes from the synthesis step itself, with the remainder from genuine model diversity.
OpenRouter disclosed that giving the panel live web access lets models surface DRACO's own grading rubric in search results, a contamination risk the company calls coincidental rather than deliberate. The fix required one configuration line to exclude the benchmark's hosting domains from the search tools, and every published number reflects that cleaned-up run.
Anthropic Suspended Fable 5 and Mythos 5 Following U.S. Export Directive
Shortly after releasing Fable 5 and Mythos 5 last week, a U.S. export control directive forced Anthropic to suspend those models for every foreign national worldwide, citing a disputed jailbreak finding. OpenRouter announced Fusion on X on June 13, positioning it as an alternative with a promise of "Fable-level intelligence at half the price."
OpenRouter Identifies Fusion Limitations for Coding and Long-Horizon Tasks
OpenRouter states that Fusion is not a full Fable replacement. DRACO skips long-horizon work, where Fable reportedly still leads. For coding, Fusion works as a tool a coding model calls selectively, not a wholesale replacement. The launch thread split roughly two-to-one positive in sentiment tracking. AI researcher Andrew Trask called it "a way bigger deal than it seems," arguing frontier labs will never again own the frontier alone. Skeptics cited bad coding results, poor tool calling, and a lack of transparency since Fable 5 is no longer available to compare results. Fusion runs entirely on models routed through OpenRouter's own infrastructure, so it does not address the export-control problem at the source.
FAQ
What did OpenRouter launch on June 12?
OpenRouter launched Fusion on June 12, a server-side API that distributes prompts to multiple AI models in parallel, then uses a judge model and synthesizer to merge responses into a unified answer.
How did Fusion's budget panel perform on the DRACO benchmark compared to Claude Fable 5?
On Perplexity's DRACO benchmark, Fusion's budget panel combining Gemini 3 Flash, Kimi K2.6, and DeepSeek V4 Pro scored 64.7%, landing within one percentage point of solo Fable 5's 65.3% score at roughly half the cost.
Why did Anthropic suspend Claude Fable 5 and Mythos 5?
Anthropuic suspended Fable 5 and Mythos 5 last week following a U.S. export control directive citing a disputed jailbreak finding, affecting access for every foreign national worldwide.
