Qvac
QVAC SDK puts on-device AI in one JavaScript or Python API — text, vision and audio with no cloud and no API keys
Pick QVAC when offline operation, data residency, or fine-tuning on the user's own hardware is the point. One API spanning text, vision and audio on iOS, Android and desktop is genuinely rare in local-first tooling. The caution is release discipline: recent 0.x versions shipped breaking changes, so pin versions and budget time for migration notes. If you need frontier-level reasoning on messy open-ended tasks, a hosted API still wins.
Verified 3h ago · liveness 74/100 · cite: rightaichoice.com/tools/qvac
- Mobile developers who need the same AI API on iOS, Android and desktop
- Apps handling sensitive documents — invoices, clinical notes — that cannot leave the device
- Offline and field deployments: planes, rural clinics, low-connectivity classrooms
- Teams that want models fine-tuned on local data via on-device LoRA
- Non-developers who want a finished chat app to open today
- Products that need frontier-model reasoning on messy open-ended tasks
- Teams unwilling to absorb breaking changes — 0.19.0, 0.20.0 and 0.21.0 all broke integrations
We scan live Reddit threads, YouTube comments, X posts, G2 reviews and other communities — and hand you an honest verdict in under a minute.
- Honest verdict, not marketing
- Real pros & cons from real users
- Attributed quotes with receipts
3 free scans · no card needed
Skip QVAC if you need frontier-model reasoning on messy open-ended tasks, or you cannot pin dependency versions — v0.19.0, v0.20.0 and v0.21.0 each shipped breaking changes inside one month.
Every release is free, but you carry the hardware cost: an 8 GB laptop or phone covers most everyday jobs, while video workloads need 32 GB.
QVAC is Apache 2.0 at $0 with no API keys, no per-token pricing and no rate limits — so the comparison is not licence cost but hardware and engineering time. Against metered hosted APIs you avoid a per-user inference bill entirely; against other local runtimes you get text, vision, audio and on-device LoRA in one API rather than a desktop-first text runtime.
In short
Qvac — QVAC SDK puts on-device AI in one JavaScript or Python API — text, vision and audio with no cloud and no API keys. Best for Mobile developers who need the same AI API on iOS, Android and desktop, Apps handling sensitive documents — invoices, clinical notes — that cannot leave the device, Offline and field deployments: planes, rural clinics, low-connectivity classrooms. Free to use.
What's new in Qvac
Checked 5 days agoAcross the latest 5 updates: 5 changelog entries.
QVAC SDK v0.21.0
QVAC SDK v0.21.0 released with breaking changes and published to npm as @qvac/sdk 0.21.0.
QVAC SDK v0.20.1
Patch release published to npm as @qvac/sdk 0.20.1.
QVAC SDK v0.20.0
QVAC SDK v0.20.0 shipped with RAG support and breaking changes, moving RAG into @qvac/cli.
QVAC SDK v0.19.1
Patch release published to npm as @qvac/sdk 0.19.1.
QVAC SDK v0.19.0
QVAC SDK v0.19.0 added RAG support with breaking changes and introduced the @qvac/rag package.
What people actually say about Qvac — is it worth it?
We ran a structured research pass across product reviews, community discussions, and post-purchase forum threads to surface the patterns vendors won't publish themselves. Below: the recurring strengths, the hidden costs people mention most, and the cohort that consistently regrets adopting this tool.
28 mentions across 4 sources (Hacker News, YouTube, GitHub, Lemmy) · researched Aug 16, 2026.
Average across the 4 sources that answered — each source counts once, not each post.
- +Fully on-device AI across Linux, macOS, Windows, Android, and iOS
- +Single API covering LLM, speech, translation, and vision tasks
- +Leader in on-device fine-tuning with 1-bit LoRA on mobile
- +Strong privacy: no data leaves the device, no API keys needed
- +P2P sharing enables decentralized, offline-capable AI apps
- −GitHub activity low: only 420 stars, 93 open issues signal immaturity
- −Steep learning curve for non-experts due to advanced concepts
- −Tether's corporate reputation creates mistrust and uncertainty
- −On-device models lag cloud alternatives for complex tasks
- −Documentation and tutorials could be more extensive for beginners
- • Potential costs for running models on mobile hardware (battery, compute)
- • No official support, only community and GitHub
- • Future monetization unclear, possibly tie-in to Tether services
Viability Score
How well maintained and how widely used is Qvac? Built from what the vendor actually publishes (docs, changelog, tutorials, integrations, pricing), whether the site is live, and how much real users discuss it. How we calculate this
Last calculated: October 2026
How we score →Key Features
- Single API for 10+ on-device AI tasks via npm install @qvac/sdk (JavaScript or Python)
- Text generation, chat, completion and structured output running locally
- On-device speech-to-text transcription
- Offline translation, including TranslatePsy-AfriSLM across 19 African languages
- Text-to-speech synthesis on-device
- Local image generation with no cloud call
- Image understanding with on-device vision models
- OCR text extraction from images, scans and photographed pages
- Image upscaling to high resolution on-device
- Embeddings plus RAG retrieval over local documents (@qvac/rag, folded into @qvac/cli)
- Fast classification with confidence scores, millisecond labels
- LoRA fine-tuning on-device, including on mobile phones, through QVAC Fabric
- Fabric runs on any GPU via Vulkan and Metal with dynamic tiling for tight mobile GPU memory
- Peer-to-peer model and workload offload between devices
- Local HTTP server with official plugins for OpenCode and OpenClaw
About Qvac
QVAC is Tether Data's Apache 2.0 platform for running AI on hardware you already own — a phone, a laptop, a robot, a wearable. The entry point is the QVAC SDK: one install (npm install @qvac/sdk) in JavaScript or Python covers 10+ on-device tasks through a single interface, spanning text generation and structured output, embeddings, RAG retrieval, speech-to-text, offline translation, text-to-speech, local image generation, OCR, vision understanding, image upscaling and millisecond classification. Nothing routes through a third-party server, so there's no API key to rotate and no per-token bill. The ecosystem splits into roles. QVAC Fabric handles hardware-agnostic LoRA fine-tuning on-device using Vulkan and Metal, with a dynamic tiling algorithm aimed at tight mobile GPU memory — including on a phone. QVAC Psy publishes open-weight models sized for devices rather than shrunk to fit, such as MedPsy for clinical reasoning and TranslatePsy-AfriSLM covering 19 African languages offline. Genesis supplies synthetic STEM reasoning data for researchers training their own small reasoning models. Vendor-listed deployments show the intended shape: Hiraia, a free offline science tutor on entry-level Android phones; Diamesh, a five-agent ophthalmology copilot on local MedPsy models; GENE.01, a humanoid fusing touch, force and vision on board. It suits mobile and desktop developers who need one AI API across iOS, Android and desktop, and teams whose data cannot leave the device. Set against hosted APIs and other local runtimes, QVAC's differentiators are its unified text/vision/audio API across desktop and mobile, peer-to-peer offload between devices, and on-device LoRA fine-tuning — but it is pre-1.0 and its release cadence has carried breaking changes.
Behind the Verdict
We'd reach for QVAC when the constraint is architectural, not cosmetic. If prompts and files legally cannot leave the device — clinical notes, invoices, field data with no signal — hosting inference somewhere else is a policy promise, while QVAC makes it a property of the build. The single text/vision/audio API across iOS, Android and desktop removes a lot of per-platform glue, and peer-to-peer offload between devices is the kind of thing general-purpose local runtimes still don't offer. The fine-tuning story is the sharper edge. Fabric runs on any GPU via Vulkan and Metal and uses dynamic tiling to fit mobile GPU memory, which means adapters can be trained where the user's data actually lives. For teams with a narrow domain — a dialect, a document type, a sensor stream — that is a meaningful step past downloading someone else's checkpoint. Where it bites: maturity. Versions 0.19.0, 0.20.0 and 0.21.0 all shipped breaking changes, and RAG moved between packages along the way (@qvac/rag, then into @qvac/cli). Plan for version pinning and migration reading, not just an afternoon integration. Pass if your product needs frontier-model reasoning on messy open-ended tasks, or if procurement requires a vendor-managed SLA and commercial support contract. This is open source with a community room (Discord, GitHub), not a managed service. Non-developers wanting a finished chat app should also look elsewhere. Closest alternative: Ollama if you want a desktop-first local runtime with a big model catalog. QVAC's bet is mobile and fine-tuning; Ollama's is the developer laptop. They overlap less than the marketing suggests.
Researching Qvac? Get your full AI stack in 60 seconds.
Free, no signup — tell us your goal and get tools matched to your budget & existing stack.
Real-world workflow fit
Concrete scenarios for the personas Qvac actually fits — and what changes day-one when you adopt it.
Installs @qvac/sdk, wires transcription, RAG over local notes and text generation into an iOS and Android app so nothing leaves the handset.
Outcome: A working on-device assistant without API keys, key rotation or a per-user inference bill as the install base grows.
Runs MedPsy or a domain model locally over sensitive records, using @qvac/rag for grounded retrieval and classification for triage labels.
Outcome: Answers and labels stay on the machine, so residency rules are satisfied by architecture rather than by a vendor policy.
Deploys an app that transcribes, translates between 19 African languages and generates images with no connection, syncing only when back in range.
Outcome: Full AI in a rural clinic, on a plane or in a low-connectivity classroom where a cloud call would simply fail.
Use Cases
- Build a private chatbot that runs entirely on a user's phone
- Create an offline voice assistant using local speech-to-text
- Develop a peer-to-peer language translation app covering 19 African languages
- Integrate an on-device AI assistant into a privacy-focused desktop application
- Turn invoice folders into accounting tables without uploading anything
- Sort documents into folders locally, with nothing leaving the machine
- Run brain-to-text models fully on-device for accessibility
- Deploy self-improving AI agents locally without cloud dependency
Models Under the Hood
as of 2026-10-17
Limitations
- QVAC runs entirely on-device, so you must download models and provision your own hardware; performance depends on local GPU support via the Fabric engine (Vulkan and Metal).
- The SDK is pre-1.0 and moving fast — the changelog shows consecutive breaking releases through v0.21.0 (Oct 4, 2026) and v0.21.1 (Oct 16, 2026), so version pinning matters for production.
- Model memory needs vary, with the vendor's own guides noting video workloads require roughly 32 GB.
- On-device models are small by design, and vendor-published benchmarks (e.g.
- Genesis III STEM results) are vendor-supplied rather than independently verified.
as of 2026-10-12
Verification history
We have re-verified Qvac 10 times since . Each pass re-reads the vendor's own pages and re-checks every listed field against that evidence; passes where nothing had changed are marked as such.
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
Showing the 6 most recent of 10 verification passes.
Free to cite with attribution — this page re-verifies continuously.
12-month cost
Project the real annual outlay, including the implied monthly cost when only an annual tier is published.
Vendor list price only. Add-on usage, seat overages, and contract minimums are surfaced under Hidden costs & gotchas.
Plans compared
For each published Qvac tier: who it actually fits, and what it adds vs. the previous tier. Cross-reference the cost calculator above for projected annual outlay.
Open source (Apache 2.0)
$0
Ideal for
Mobile, desktop and robotics developers, plus researchers who want on-device inference and LoRA fine-tuning without per-token billing or a cloud dependency
What this tier adds
Starting tier: the full QVAC SDK, Fabric engine, QVAC Psy open-weight models, Genesis dataset for research and non-commercial use, and peer-to-peer offload — all at $0 with no API keys or rate limits
Where the pricing makes sense
The company stage and team size where Qvac's pricing actually pencils out — and where peers do it cheaper.
QVAC is Apache 2.0 at $0 with no API keys, no per-token pricing and no rate limits — so the comparison is not licence cost but hardware and engineering time. Against metered hosted APIs you avoid a per-user inference bill entirely; against other local runtimes you get text, vision, audio and on-device LoRA in one API rather than a desktop-first text runtime.
Setup time & first value
How long it actually takes to get something useful out of Qvac — broken out by persona, not the marketing-page minute.
Developers: minutes to a first local completion after npm install @qvac/sdk; a first useful feature lands the same day, with model download time depending on your connection and disk. Teams adding RAG or Fabric fine-tuning: budget days, because you provision local GPU support and pin around breaking releases. Non-developers: not the target — there is no finished chat app to open.
Switching to or from Qvac
How to bring data in from common predecessors and how to get it back out — written for the switcher, not the buyer.
- →From a hosted inference API: replace per-call endpoints with the QVAC SDK task interface, then add @qvac/rag for retrieval you previously hosted.
- →From a desktop-first local runtime: port text prompts to the QVAC SDK and gain transcription, vision and image generation through the same interface.
- →From a larger cloud model: adopt a QVAC Psy open-weight model sized for the device, and use Fabric to fine-tune on your own data.
- →From a scripted prototype: move to @qvac/cli, which now hosts RAG alongside the SDK tooling.
- ↗To a hosted frontier API: keep your prompt layer and swap the QVAC task calls for remote endpoints where reasoning quality matters more than locality.
- ↗To Ollama: if your target is desktop-first text generation rather than mobile and multimodal, a desktop-focused runtime is the simpler fit.
- ↗To a managed pipeline: move RAG and orchestration to a hosted vector store and pipeline service if you no longer need data to stay on-device.
Integrations
Resources & Guides
Tutorials & Learning
YouTube returned 6 videos for “Qvac”, and we withheld 6: 6 could not be judged, because “Qvac” is a single word that other videos use for other things. We are showing none, because we could not prove any of them are about Qvac.
Official links
Tools that pair well with Qvac
Common stack mates teams adopt alongside Qvac, with the specific reason each pairing earns its keep.
LLM Hub
Run 15+ offline AI models on your phone for private, on-device chat, image, music and code generation.
Cactus
Cactus runs Needle on-device inference in sub-30MB binaries on phones, wearables, and Raspberry Pi, with cloud fallback.
LiveKit
Build, test, and scale real-time voice, video, and robotics agents on an open-source framework with published per-minute inference pricing.
Featured Head-to-Head Comparisons
Qvac vs Spider Cloud
Choose Spider Cloud if you need high-scale, real-time web data for AI agents or RAG pipelines with a pay-as-you-go model. Choose Qvac if you prioritize data privacy, offline operation, and on-device inference across mobile and desktop—and you're willing to trade cloud capabilities for complete local control.
Qvac vs Temporal Ai
Choose Temporal AI if you need durable, fault-tolerant orchestration for AI agents and microservices that survive failures and provide full execution visibility; it's the standard for reliability at scale. Choose Qvac if you prioritize privacy, offline operation, and on-device AI (LLMs, speech, translation) across mobile and desktop without cloud dependency. They solve fundamentally different problems.
Qvac vs Voyage Ai
Choose Voyage AI if you need production-grade embeddings/rerankers for enterprise RAG, especially on domain-specific (finance, legal) data with long-context support. Choose Qvac if you are a developer building a privacy-focused, offline, cross-platform app that requires local AI inference without cloud dependency. They serve completely different needs.
Alternatives to Qvac
View all
Frequently Asked Questions
Used Qvac? Help shape our editorial sentiment research.