Primordial is a proprietary SDK that gives you on-device AI application infrastructure, so you can build AI features without token billing, privacy concerns or cloud APIs. Primordial is a proprietary SDK that gives you on-device AI application infrastructure, so you build AI features without token billing, privacy concerns or cloud APIs.
On-device AI lets you ship with predictable pricing without token billing, usage caps, or API bills.
On-device AI lets you ship with predictable pricing without token billing, usage caps, or API bills.
Expected Cost Savings
$42,000/yearAt 1M requests per month.
GPT-5.4 · Assumes 500 input and 150 output tokens per request. · SourceExpected Cost Savings
$13,315/yearBring AI to apps and workflows without an internet connection with no cloud server round-trip latency while keeping user data on-device by default.
These estimates don’t even include the cost of building, scaling, monitoring, and maintaining the system.
These estimates don’t even include the cost of building, scaling, monitoring, and maintaining the system.
More efficient, faster chips and models now make useful intelligence possible inside everyday devices.
State of smartphone and edge AI chips
Google and Apple already including AI
Move beyond chat demos. One SDK for the fastest path from on-device intelligence to production apps and enterprise workflows.
Manage availability, background downloads, loading, shared access, memory pressure, and recovery.
Build local collections and retrieve the right context by meaning, not keywords.
Generate answers from your data and return citations to the sources used.
Summarize, extract, classify, or define a generation task for your product.
Create model-only or grounded conversations with local context and citations.
Record and transcribe audio locally with managed session and model readiness.
Record speech, transcribe it, and summarize the transcript into three bullet points in a few lines of code.
Import and activate Primordial with a client instance and your evaluation key.
import Primordial
let primordial = PrimordialClient()
try await primordial.activate(.evaluationKey("pk_eval_your_key_here"))Download and verify the configured speech and text generation models.
for try await progress in primordial.ai.makeAvailable([
.voiceInput,
.generation
]) {
print(progress.fractionCompleted ?? 0)
}Record speech and collect the local transcript after the user stops recording.
let recording = primordial.recording()
try await recording.start()
let result = try await recording.stop()Summarize the transcript into three bullet points.
let summary = try await primordial.summarize(
result.transcription.text,
style: .bulletPoints(maximum: 3)
)
print(summary)import Primordial
let primordial = PrimordialClient()
try await primordial.activate(.evaluationKey("pk_eval_your_key_here"))
for try await progress in primordial.ai.makeAvailable([
.voiceInput,
.generation
]) {
print(progress.fractionCompleted ?? 0)
}
let recording = primordial.recording()
try await recording.start()
let result = try await recording.stop()
let summary = try await primordial.summarize(
result.transcription.text,
style: .bulletPoints(maximum: 3)
)
print(summary)