Synth appears within TrustedRouter's documentation as a family of versioned presets rather than a single monolithic model, with named aliases such as iris-2.0, prometheus-2.0, and zeus-1.0 exposed for direct routing alongside code-tuned variants like iris-code-1.0, prometheus-code-1.0, and zeus-code-1.0. This preset-oriented design lets users select a specific Synth "flavor" through the trustedrouter/synth model key while the underlying engine swaps behind the scenes, a pattern that treats Synth as a configurable routing tier rather than a fixed-weights release. The framing in the gateway docs positions Synth presets as tunable compositions, with code-specialized and general-purpose variants both available so that chat, reasoning, and programming workloads can each be routed to a more targeted configuration.
In practical terms, choosing trustedrouter/synth through the OpenAI-compatible endpoint gives a developer access to that preset family without managing multiple vendor SDKs, and the same docs pair Synth presets with related advisor and synthesis routes such as trustedrouter/socrates and the openpatcher compositions that mix different advisors. Because the gateway is the only source describing Synth, the qualitative story centers on routing flexibility: users pick a Synth preset to influence reasoning style and tool-use behavior, then layer privacy controls like zdr or confidential compute on top. This makes Synth a practical fit for teams that want a single integration point while still being able to tune between general reasoning, advised synthesis, and code-leaning variants as workloads change.
