OrcaRouter, built by Singapore-based Continuum AI Pte. Ltd., has launched Routing DSL, a programmable routing framework positioned as a Fable 5 alternative for developers who lost access to Claude Fable 5 under recent export-control restrictions. Rather than swapping in a single replacement model, Routing DSL lets teams define routing logic in YAML and CEL expressions to dynamically orchestrate 200+ models based on prompt complexity, task type, latency, cost, and safety policies \u2014 routing simple requests to efficient models, escalating hard prompts to frontier models, running models in parallel and merging the best output, and applying guardrails and fallback chains. Built into OrcaRouter’s OpenAI-compatible AI Gateway, it delivers roughly 40% lower cost, sub-1ms routing overhead, and no token markup, with early internal evaluations suggesting well-designed strategies can reach Fable 5\u2013level performance at a fraction of the cost. The launch reframes the core question from “Which model?” to “Which strategy?”, presenting routing as a new programmable control layer in the AI stack. Routing DSL is available immediately to all OrcaRouter users.<\/p>\n
Since \u201cBetter \u201cWhen For Unlike \u2022\u00a0\u00a0\u00a0\u00a0 Route simple requests to \u2022\u00a0\u00a0\u00a0\u00a0 Automatically escalate \u2022\u00a0\u00a0\u00a0\u00a0 Run multiple models in \u2022\u00a0\u00a0\u00a0\u00a0 Configure fallback chains \u2022\u00a0\u00a0\u00a0\u00a0 Apply guardrails and \u2022\u00a0\u00a0\u00a0\u00a0 Optimize for cost, Routing DSL arbiters: Routing Instead Blended cost versus As Combined Routing Documentation: <\/b>docs.orcarouter.ai\/routing\/routing-dsl<\/a><\/p>\n OrcaRouter OrcaRouter OrcaRouter Learn more: <\/b>orcarouter.ai<\/a><\/p>\n
\naccess to Claude Fable 5 was suspended under a recent export-control directive,
\nmany developers have been searching for \u201cFable 5 banned\u201d and for a reliable
\nFable 5 alternative. Today, <\/a>OrcaRouter<\/a> delivers that answer. The company has
\nunveiled Routing
\nDSL<\/a>,
\na programmable routing framework that lets developers define in code exactly
\nhow each AI request is evaluated, routed, combined, and optimized across 200+
\nmodels. Early internal evaluations show that well-designed strategies can match
\nfrontier models like Claude Fable 5 \u2014 making OrcaRouter a practical Fable 5
\nalternative for teams that no longer have direct access to the model itself.<\/p>\n
\nintelligence comes from orchestration \u2014 not from bigger models alone.\u201d<\/i><\/b><\/p>\n
\naccess to Fable 5 was suspended, teams didn\u2019t just lose a model \u2014 they lost an
\nentire capability tier,\u201d said [Title] [Spokesperson Name] of OrcaRouter.
\n\u201cRouting DSL gives that tier back. Instead of a single fixed Fable 5
\nalternative, you get a strategy: compose specialized models, run them in
\nparallel, escalate only the hard prompts, and reach Fable 5\u2013level results at a
\nfraction of the cost.\u201d<\/p>\nWhy developers are searching for a Fable 5 alternative<\/h1>\n
\nmany teams, \u201cFable 5 banned\u201d wasn\u2019t a headline \u2014 it was a production incident.
\nWorkloads built around Claude Fable 5 suddenly needed a Fable 5 alternative
\nthat matched the quality, without months of re-engineering. Routing DSL is
\nbuilt for exactly this moment: instead of swapping one model for another, it
\nlets developers compose Fable 5\u2013level performance from the 200+ models that are
\nstill available.<\/p>\nRouting as code, not configuration<\/h1>\n
\nstatic model selection, Routing DSL lets teams define their routing logic with
\nsimple YAML and CEL expressions, choosing models dynamically based on prompt
\ncomplexity, task type, latency, cost targets, safety policies, and custom
\nbusiness rules. In just a few lines of configuration, developers can:<\/p>\n
\nefficient open-source models<\/p>\n
\nhard prompts to frontier models<\/p>\n
\nparallel and merge the best result<\/p>\n
\nand reliability policies<\/p>\n
\ngovernance rules before execution<\/p>\n
\nquality, latency, or custom objectives<\/p>\n
\nhow the outputs of multiple models are merged into a single answer<\/i><\/p>\n
\nDSL is built directly into OrcaRouter\u2019s AI Gateway and supports 200+ leading
\nmodels through a single OpenAI-compatible endpoint \u2014 with roughly 40% lower
\ncost, sub-1ms routing overhead, and no token markup.<\/b><\/p>\nFable 5\u2013level intelligence through orchestration<\/h1>\n
\nof paying for a single expensive model on every request, Routing DSL lets
\ndevelopers combine specialized models and parallel execution, allocating
\ncompute only where it delivers measurable quality gains. Early internal
\nevaluations indicate that carefully designed strategies can reach Claude Fable
\n5\u2013level performance at a fraction of the cost \u2014 the closest thing to Fable 5
\nfor the many developers who can no longer access the model directly. The
\nquestion shifts from \u201cWhich model?\u201d to \u201cWhich strategy?\u201d<\/p>\n
\nrunning the full panel on every request, by share of hard prompts<\/i><\/p>\nA new layer in the AI stack<\/h1>\n
\nAI systems become more agentic, routing is no longer just a setup step \u2014 it is
\nbecoming application logic in its own right. Routing DSL provides a
\nprogrammable control plane for AI workloads: a single place to define how
\nintelligence is composed, governed, and optimized in production.<\/p>\n
\nwith OrcaRouter\u2019s adaptive routing engine, observability platform, guardrails,
\ngovernance controls, and Agent Firewall, Routing DSL gives organizations a
\nunified way to build reliable, cost-effective AI systems at scale \u2014 and a
\nflexible Fable 5 alternative that doesn\u2019t rely on any single restricted model.<\/p>\nAvailability<\/h1>\n
\nDSL is available immediately to all OrcaRouter users.<\/p>\n
\nFable 5 Fusion API<\/a><\/p>\n
\nFable 5 Fusion Mini API<\/a><\/p>\n
\nFable 5 Fusion Flash API<\/a><\/p>\n