Pricing
1% take rate. Lowest price possible.
99% of every dollar goes to the provider running the job. Quotes are locked at submission. We're the cheapest in the market on the workloads Apple Silicon was built for.
Apple-Silicon-native workloads
Apple Neural Engine, Apple Vision, VideoToolbox, Translation, MLX. Hardware NVIDIA clouds don't have. We're the cheapest option in the market on every row below.
Workload
Unit
Batch
Realtime
Compare against
Savings
OCR (Apple Vision)Soon
per 1k pages
$0.050
$0.10
AWS Textract @ $1.50
−97%
Background removalSoon
per image
$0.0005
$0.0010
remove.bg @ $0.20
−99%
Image upscale (2×/4×)Soon
per image
$0.0010
$0.0020
Replicate Real-ESRGAN @ $0.005
−80%
Aesthetics scoringSoon
per image
$0.0003
$0.0006
manual review
—
Transcription (Whisper)
per audio hour
$0.13
$0.26
OpenAI Whisper API @ $0.36
−64%
Subtitle generationSoon
per video hour
$0.15
$0.30
AssemblyAI @ $0.45
−67%
Audio cleanupSoon
per audio hour
$0.067
$0.13
Krisp Enterprise (subscription)
—
Text-to-speechSoon
per 1k chars
$0.0050
$0.010
ElevenLabs @ $0.30
−98%
TranslationSoon
per 1M chars
$0.10
$0.20
Google Translate @ $20
−99%
Embeddings
per 1M tokens
$0.0088
$0.018
OpenAI text-embedding-3-small @ $0.02
−56%
Video transcode
per output minute
$0.011
$0.022
AWS MediaConvert @ $0.017
−35%
HEIC/HEVC conversionSoon
per GB
$0.0050
$0.010
AWS MediaConvert @ ~$0.017/min
—
Image generation
per image
$0.0022
$0.0044
Together FLUX-schnell @ $0.0027
−19%
Xcode build farmSoon
per build minute
$0.022
$0.044
Xcode Cloud @ $0.60/hr (no minimum)
−39%
iOS test farm (XCTest)Soon
per test minute
$0.022
$0.044
Sauce Labs @ ~$1/min
−97%
Prices in USD. Batch tolerates 3–6 sec dispatch. Realtime is 2× batch (sub-second). Quotes locked at submission, capped by max_spend_usd. Failed tasks aren't billed.
Estimator
How much would your workload cost?
Workload
Priority
Volume
10 audio hour
Estimated cost
$1.30
Transcription (Whisper) · batch · 10 audio hour
Unit price$0.130 / audio hour
vs. hyperscalersave 64%
Also available
H100-class hardware wins on price for these workloads. We support them so you can hit one API for everything — but if price is the only thing that matters, route them to a hyperscale inference host.
Workload
Unit
Batch
Realtime
Compare against
Savings
Chat (8B class)
per 1M tokens
$8.80
$17.60
Together Llama 3.1 8B @ $0.09
—
Long-context LLM (70B)
per 1M tokens
$44.00
$88.00
Together Llama 3.3 70B @ $0.44
—
Image captioning
per 1k images
$2.20
$4.40
GPT-4o vision @ ~$0.50
—
Stable Diffusion (SDXL)Soon
per image
$0.011
$0.022
Replicate SDXL @ $0.0055
—
PyTorch (MPS)Soon
per GPU minute
$0.022
$0.044
RunPod A100 Spot @ $0.013/min
—
Blender (Metal)Soon
per render minute
$0.044
$0.088
AWS Deadline + Spot @ ~$0.03/min
—
LoRA fine-tuneSoon
per training minute
$0.022
$0.044
OpenAI fine-tune @ $25/1M tokens
—
Small LLM (private RAG)Soon
per 1M tokens
$3.70
$7.40
Together Llama 8B @ $0.18
—
Custom batchSoon
quoted
—
—
Bring-your-own container
—
Same per-task contract as featured workloads — locked quote, max_spend cap, no charges on failure. Pricing reflects honest provider cost on Apple Silicon, not a market-matched price.
FAQ