Pin AI music model ids: music_v2_5, v6, lyria-3.5 and sume/music-auto
Which model id to pin for ElevenLabs Music, Suno and Google Lyria, and what Sume does with sume/music-auto. A config table and a rule for logging the engine.

Pin the model id in your music config instead of relying on a vendor default, and log the engine that actually ran. ElevenLabs lists music_v2_5, music_v2 and music_v1 in its API docs; Suno names v6, v6-wild and v6-mini; Google lists lyria-3.5. On Sume, sume/music-auto lets Sume pick (Lyria 3.5 today) and lyria-3.5 pins it; the job records the engine in job.request.routed_model.
The ids, by vendor
ElevenLabs made Music v2.5 the default for prompted and reference generation, and its announcement says Music v2 remains available. The API docs page lists the API ids. Suno's help page names its v6 models, and the v6 post says previous models will be retired. Google's pricing page lists the Lyria ids.
| Vendor | Identifiers | Notes |
|---|---|---|
| ElevenLabs | music_v2_5, music_v2, music_v1 | v2.5 is described as the most advanced; v1 stays during a transition period |
| Suno | v6, v6-wild, v6-mini | Suno says it will retire previous models (no date given); v6 and v6-wild are Pro or Premier only |
| lyria-3.5, lyria-3-clip-preview, lyria-3-pro-preview | Preview names may change | |
| Sume Music Router | sume/music-auto, lyria-3.5, lyria-3-pro | Unknown id gives 400 model_not_found |
Rules for a stable integration
A default is a moving target: ElevenLabs made v2.5 the ElevenMusic default on September 11, 2026, and Suno plans to retire earlier models. A pipeline that omits the model silently changes behaviour when the vendor flips it.
- Always send an explicit model id where the vendor allows it.
- Store the id you requested and the id that ran.
- Treat "preview" ids as short-lived and re-check them monthly.
- Re-audition a brief when the default changes, rather than assuming equivalence.
What this looks like on Sume
The Music Router docs say that job.model echoes the requested id and that job.request.routed_model names the engine that ran, so the two fields together give you the "requested vs ran" pair. Pinning lyria-3.5 makes a campaign repeatable at the engine level, though results still vary between calls because there is no seed. GET /v1/music-router/models returns the catalog, so validate your config against it rather than against a blog table.
Write it down
Put the id table in your repository and date it. When a vendor retires a model or changes a default, you will know which campaigns used it. The Music 1.0 page notes that Music 1.0 routes now resolve through the router, so older integrations can keep sume/music-1.0 while new ones call the router. Poll and fetch jobs as described in Jobs and results.
Sources
Related posts
More in Developers
- 1080x1350 sent as aspect_ratio: what Ideogram, Grok, Imagen get
Send pixels instead of a ratio and Sume snaps to the nearest native ratio. Tested table for 1080x1350, 1200x628 and 1500x500 across four model families.
- Poll Sume audio jobs at the recommended interval, not a fixed sleep
The Sume job status response tells you when to poll next and whether the job is terminal. Read those fields instead of hardcoding sleep(2).
- Ported Sora wrapper blocked until done? Sume sync stops at 30 s
A wrapper that blocks until the video is done will time out on Sume sync mode, which waits at most 30 seconds. Return the job id and poll, never resubmit.
- Pre-flight tool access: tools_list on Sume's MCP
Notion added a tool that reveals connection-scoped capabilities before requests. Sume's MCP does the same job with tools_list, tools_schema and mcp_health.
Written by Sume