actaserve

Simple ways to
start building.

Use the model API at published rates. Reserve dedicated compute with a quote shaped around your workload.

Model API

Usage-based pricing

Connect through OpenAI- and Anthropic-compatible interfaces. The active model catalogue lists current rates and billing units.

  • Choose from the active model catalogue
  • Published pricing for each model
  • $5 in account credit after email verification
Get an API key View model rates

Model capabilities and pricing apply. API access does not reserve dedicated hardware or guarantee data residency.

Dedicated compute

Custom quote

Reservations open

Actaserve will own and operate the hardware. Choose root/bare-metal access or a managed API endpoint.

  • Hardware procurement, power and cooling
  • Model bring-up and ongoing operations
  • Capacity, location and access scoped per project
Reserve capacity

Dedicated capacity is reserved per contract, with the delivery date set in your quote. It is not ready-to-use inventory. Deployment terms.

Find your model. See your rate.

Published prices in USD. Token prices below are per one million tokens; other models use their listed billing unit. Check the model's supported options before sending a request.

533 models

Active public model prices in US dollars
ModelTypeInput / 1M tokensOutput / 1M tokensOther billing unit
Meta: Llama 4 Scoutmeta-llama/llama-4-scout Language $0.1$0.3Token based
Meta: Llama Guard 4 12Bmeta-llama/llama-guard-4-12b Language $0.18$0.18Token based
Meta: Muse Glimmer 30Bmeta/muse-glimmer-30b Language $0.3$1.2Token based
Meta: Muse Spark 1.1meta/muse-spark-1.1 Language $1.25$4.25Token based
Meta: Muse Spark 1.2meta/muse-spark-1.2 Language $1.25$4.25Token based
Meta: Muse Spark 1.2 Contributormeta/muse-spark-1.2-contributor Language $0.1$0.2Token based
Meta: Muse Spark 1.3meta/muse-spark-1.3 Language $1.25$4.25Token based
Meta: Muse Spark 1.3 Contributormeta/muse-spark-1.3-contributor Language $0.1$0.2Token based
Microsoft: Phi 4microsoft/phi-4 Language $0.07$0.14Token based
MiniMax: MiniMax-01minimax/minimax-01 Language $0.2$1.1Token based
MiniMax: MiniMax M1minimax/minimax-m1 Language $0.55$2.2Token based
MiniMax: MiniMax M2minimax/minimax-m2 Language $0.3$1.2Token based
MiniMax: MiniMax M2.1minimax/minimax-m2.1 Language $0.3$1.2Token based
MiniMax: MiniMax M2.5minimax/minimax-m2.5 Language $0.27$1.08Token based
MiniMax: MiniMax M2.7minimax/minimax-m2.7 Language $0.21$0.84Token based
MiniMax: MiniMax M2-herminimax/minimax-m2-her Language $0.3$1.2Token based
MiniMax: MiniMax M3minimax/minimax-m3 Language $0.3$1.2Token based
Mistral: Codestral 2508mistralai/codestral-2508 Language $0.3$0.9Token based
Mistral: Codestral 2508 (batch)mistralai/codestral-2508:batch Language $0.15$0.45Token based
Mistral: Devstral 2 2512mistralai/devstral-2512 Language $0.4$2.0Token based

Reported cached input tokens are billed at 10% of the published input rate. Model availability can change. Read the quickstart or open the catalog API (JSON).

Bring your own model.

A dedicated-compute option, custom quoted

Deploy private weights on Actaserve-owned infrastructure, subject to model, licence and runtime qualification. Bring your checkpoint and serving requirements so we can scope model bring-up.

Reserve capacity

Ready for your first call?

Explore the docs