Dedicated deployment offered by Together’s model card. Provision GPU capacity; not a current serverless listing. Pricing includes both character usage and GPU time.
Regions
Dedicated deployment region and capacity depend on the provisioned endpoint.
Request mode
Together speech API supports synchronous, HTTP streaming and WebSocket delivery; use the dedicated endpoint configuration and host-specific schema.Source for Request mode (opens in a new tab)
The announcement explicitly distinguishes bilingual Turbo from eleven-language v3. Turbo’s card also contains conflicting eleven-language v3 copy; confirm the deployed variant.
Prices retain their published currencies and billing units below.
Published billing units · original currency · observed 13 Oct 2026
Category
Usage
Rate
Unit
Speech synthesis
Speech license; GPU billed separately1
$10.00
1M characters
Rate conditions
This $10/1M-character component excludes required GPU hourly charges, whose amount depends on hardware and provisioning. It is not an all-inclusive serverless rate. Native Rime cloud retirement does not automatically retire this dedicated deployment.
General Together policy stores prompts/responses by default with an organization ZDR setting; training sharing is separately opt-in. The dedicated Arcana launch advertises ZDR, so confirm the actual dedicated endpoint contract and logging configuration rather than applying a blanket assumption.
Caching can reuse earlier input processing. Availability, lifetime and billing depend on the endpoint. A cache hit still generates a new response.
01 / INITIAL INPUTProcess & retain
The system processes eligible input and retains reusable state.
02 / LATER REQUESTCheck available state
Reuse depends on matching input, session state and the cache policy.
03 / CACHE HITReuse & generate
Reusing state can reduce repeated work. Any charges follow the billing rules below.
Together AI
rime-labs/rime-arcana-v3-turbo
Not verified
No model-specific prompt-cache contract or discounted cached-character rate established. TensorRT engine caches, persistent sockets and speaker profiles are not prompt caching.
How it works
Not collected
No model-specific prompt-cache contract or discounted cached-character rate established. TensorRT engine caches, persistent sockets and speaker profiles are not prompt caching.
Minimum reusable input
Not collected
No model-specific prompt-cache contract or discounted cached-character rate established. TensorRT engine caches, persistent sockets and speaker profiles are not prompt caching.
Lifetime & refresh
Not collected
No model-specific prompt-cache contract or discounted cached-character rate established. TensorRT engine caches, persistent sockets and speaker profiles are not prompt caching.
How to enable it
Not collected
No model-specific prompt-cache contract or discounted cached-character rate established. TensorRT engine caches, persistent sockets and speaker profiles are not prompt caching.
Usage fields
Not collected
No model-specific prompt-cache contract or discounted cached-character rate established. TensorRT engine caches, persistent sockets and speaker profiles are not prompt caching.
Billing rules
Not collected
No model-specific prompt-cache contract or discounted cached-character rate established. TensorRT engine caches, persistent sockets and speaker profiles are not prompt caching.
Limits & data boundaries
Not collected
No model-specific prompt-cache contract or discounted cached-character rate established. TensorRT engine caches, persistent sockets and speaker profiles are not prompt caching.
No verified cache rate for this endpoint. An empty rate is not a zero charge.
What the measurements say
No benchmark observations have been added for this model. We do not turn a vendor description into a performance score.
Latency and throughput depend on the host, request size, reasoning settings and load. There is no measured speed or global usage figure in this record.
The record behind the model
Why it is included
Rime reserves ownership of its AI voice models and grants a limited service license. Current Coda/Mist deployments explicitly use private, closed-source licensed engine images; Together describes Arcana as proprietary. The separately released Rimecaster speaker-embedding model is CC-BY-4.0 and is excluded.
