Primary sources last checked on 13 Nov 2026. Verification freshness is tracked separately from factual change, and unknown values remain explicit rather than inferred.
NotableDistinct official first-party checkpoint with substantive specialist value; publisher rankings do not establish Leading.Curation method →
Open Model Weights records ZONOS3-Mini as a Pending verification-parameter text to speech open-weight model in the ZONOS3 family, with Zyphra listed as the developer. The checked source repository provides 109 recognized weight artifacts, format evidence for GGUF, PyTorch, Safetensors and a context window that is not declared in the checked standard sources. Values that are not evidenced in the checked sources remain explicitly undisclosed or unknown.
The repository declares apache-2.0. Commercial use allowed under apache-2.0 conditions. Open Model Weights reports the checked license evidence and classification; this is not legal advice.
No base model is declared in the checked standard lineage metadata, so Open Model Weights does not infer one. Training recipe status: not-disclosed-in-checked-repository; data disclosure status: not-disclosed-in-standard-metadata-or-obvious-model-card-section.
ZONOS3-Mini is Zyphra's locally deployable multilingual text-to-speech model with a publisher-described 800M total and roughly 100M active parameters. The official Apache-2.0 release includes multiple speech runtimes and optional voice conditioning. Open Model Weights independently inspected the exact repository revision, listed weight artifacts and declared license; it has not run the model or reproduced publisher benchmark results. The source verifier found a model-card and repository license declaration, not a separately checked standard license file.
Source evidence 1The official technical card declares about 800M total and 100M active parameters, a sparse speech model that supports 43 languages according to Zyphra.Source ↗
Source evidence 2Original source components include voice and audio codec assets; GGUF/WebGPU directories and speech processing dependencies are not interchangeable with a single text-model safetensors file.Source ↗
Source evidence 3Apache-2.0 is declared for the official checkpoint; voice cloning must still be used with permission and applicable voice, likeness and data-protection safeguards.Source ↗
Source evidence 4Phone, laptop and browser operation, latency and voice quality are publisher-reported until measured on representative hardware and multilingual speech samples.Source ↗
Official immutable revision24f39592cf5521d675d96db41fc379af3ad369beSource ↗
Model license evidenceapache-2.0; declared-metadata-checkedSource ↗
Commercial-use statusallowed; see exact model-specific termsSource ↗
Repository cardREADME and source-described usage; no independent OMW inference runSource ↗
Performance provenancePublisher metrics, if mentioned, are not independent OMW measurementsSource ↗
WHY IT MATTERS
ZONOS3-Mini targets spoken audio generation and controllable voice delivery, with multilingual text input and optional reference voice conditioning.
The official model weights carry Apache-2.0 metadata. This simplifies many deployment scenarios compared with custom model licenses, subject to notice obligations.
Zyphra publishes language coverage, sample demonstrations and speed measurements, including throughput figures under particular server settings.
MODEL-SELECTION VIEW
Strong fit when…
ZONOS3-Mini targets spoken audio generation and controllable voice delivery, with multilingual text input and optional reference voice conditioning.
Notable status reflects the distinctive compact active speech footprint and multiple distribution formats, not an OMW-assessed ranking for voice realism.
The official technical card declares about 800M total and 100M active parameters, a sparse speech model that supports 43 languages according to Zyphra.
Important considerations
The advertised 100M active parameters refers to sparse routing, not the complete stored model; including vocoder and speaker conditioning changes real deployment requirements.
Apache licensing does not provide consent from speakers. Voice cloning introduces impersonation, consent, biometric privacy and misuse concerns that remain outside the repository license.
Real-time claims should be retested on target hardware with startup downloads, device memory, audio codec processing and concurrency included in the evaluation.
Purpose, modality and checkpoint
ZONOS3-Mini targets spoken audio generation and controllable voice delivery, with multilingual text input and optional reference voice conditioning.
Zyphra lists 43 languages and word-level style controls. Such coverage is a declaration, not proof of equivalent pronunciation, intelligibility or accents across all target languages.
Original weight artifacts and version identity
The first-party release contains model, codec, voice and runtime assets. Use the Zyphra repository rather than treating an unrelated GGUF or MLX mirror as the canonical source.
The advertised 100M active parameters refers to sparse routing, not the complete stored model; including vocoder and speaker conditioning changes real deployment requirements.
License terms and commercial-use limits
The official model weights carry Apache-2.0 metadata. This simplifies many deployment scenarios compared with custom model licenses, subject to notice obligations.
Apache licensing does not provide consent from speakers. Voice cloning introduces impersonation, consent, biometric privacy and misuse concerns that remain outside the repository license.
Publisher evaluations versus independent evidence
Zyphra publishes language coverage, sample demonstrations and speed measurements, including throughput figures under particular server settings.
Those figures are producer-provided. Independent comparisons should measure word error rate, speaker similarity, latency, intelligibility and robustness across target languages.
Runtime integration and resource requirements
The release describes CUDA, MLX, C++/GGUF and WebGPU options. Choose the assets and runtime matching the target device instead of assuming all backends are equivalent.
Real-time claims should be retested on target hardware with startup downloads, device memory, audio codec processing and concurrency included in the evaluation.
Practical selection and reproducibility
Notable status reflects the distinctive compact active speech footprint and multiple distribution formats, not an OMW-assessed ranking for voice realism.
Compare against source-verified Phonon variants with the exact license of each variant, multilingual coverage and an audited voice-consent workflow.
IMPLEMENTATION & EVALUATION
How to evaluate ZONOS3-Mini before deployment
Reproducible operator checklist; these steps are not claims of an OMW model run.
STEP 01
Pin original source and weight manifest
A named model does not guarantee identical checkpoint revisions.
REPRODUCIBLE ACCEPTANCE TEST
Record 40-character SHA, original file names, configuration and tokenizer/processor versions.
WHAT CAN GO WRONG
Unpinned revisions can silently change weights or usage requirements.
STEP 02
Reproduce task-specific inputs
A specialist checkpoint can fail outside the publisher's demonstrated conditions.
REPRODUCIBLE ACCEPTANCE TEST
Document actual prompt templates, modality preprocessing, inference backend and compatible software versions.
WHAT CAN GO WRONG
Using unsupported serving backends can invalidate comparisons.
STEP 03
Audit license and operational controls
Model card labels and code licenses are not automatically permission for every deployed use.
REPRODUCIBLE ACCEPTANCE TEST
Check exact weight terms, notices, tool permissions, data privacy and runtime isolation.
WHAT CAN GO WRONG
Restricted weight rights or unsafe action tools can make a deployment unsuitable.
STEP 04
Publish independent results only if measured
Vendor benchmark figures are useful evidence, but are not OMW-run observations.
REPRODUCIBLE ACCEPTANCE TEST
Pre-register held-out tasks and compare baseline checkpoint, hardware, quality, memory and error rates.
WHAT CAN GO WRONG
Attributing first-party results to independent inference would mislead readers.
The official technical card declares about 800M total and 100M active parameters, a sparse speech model that supports 43 languages according to Zyphra.
Are the original ZONOS3-Mini weights downloadable?
The official publisher repository is https://huggingface.co/Zyphra/ZONOS3-Mini. Download availability, original weight artifact names, and access rights are checked at the pinned source revision; download is not proof of deployment permission.
What license applies to ZONOS3-Mini?
Apache-2.0 is declared for the official checkpoint; voice cloning must still be used with permission and applicable voice, likeness and data-protection safeguards.
Is ZONOS3-Mini independently benchmarked by Open Model Weights?
No. Apache-2.0 is declared for the official checkpoint; voice cloning must still be used with permission and applicable voice, likeness and data-protection safeguards. Publisher tests and third-party claims are never presented as OMW-run inference or performance measurements.
Can ZONOS3-Mini run locally?
Phone, laptop and browser operation, latency and voice quality are publisher-reported until measured on representative hardware and multilingual speech samples.
Why is ZONOS3-Mini classified Notable?
The official source-released weights and distinctive technical scope justify editorial attention. Notable is not a universal ranking; Leading would require a relevant independent, checkpoint-matched evaluation.
03 · PRACTICAL TECHNICAL PROFILE
Deployment, licensing and reproducible evidence.
Inspect weight-memory estimates, model lineage and runtime evidence separately from promotional or untested performance claims.
DEPLOYMENT PROFILE
Weight memory, without pretending it is a deployment guarantee.
These bars visualize weight-only decimal-GB estimates. They exclude KV cache, activations, runtime overhead, optimizer state and sharding.
WEIGHT-ONLY MEMORY
BF16 / FP16—
FP8 / INT8—
INT4—
FIELD EVIDENCE MAP
WeightsVerified
LicenseVerified
Commercial useAllowed
ContextNot declared
FormatsVerified
LineageNot declared
Training recipeNot disclosed in checked repository
Training dataNot disclosed in standard metadata or obvious model card section
RuntimeSource-derived · not runtime-tested
LICENSE & USE
apache-2.0
Commercial use allowed under apache-2.0 conditions
Declared licenseapache-2.0
ClassificationAllowed
VerificationDeclared metadata checked
This section summarizes the checked repository evidence. It is not legal advice.
LINEAGE & PROVENANCE
What the checked sources disclose — and what they do not.
Open Model Weights does not infer a hidden base model or training corpus when the checked sources do not declare one.
LINEAGE
BASE / PARENTNot declared in model-card metadataUnknown
not inferred
CURRENT RECORDZONOS3-MiniZONOS3
TRAINING DISCLOSURE
Training recipeNot disclosed in checked repository
Training dataNot disclosed in standard metadata or obvious model card section
FamilyZONOS3
RUNTIME SIGNALS
Compatibility claims stay attached to their evidence level.
Declared and model-card-mentioned runtimes are discovery signals. They are not treated as Open Model Weights runtime tests unless explicitly measured.
Declaredzonos3
MentionedMLX
MentionedNeMo
MentionedSGLang
OMW runtime testNot performed
EVIDENCE TRAIL
Sources and observed history.
This is where the article remains auditable: source links on one side, the registry's observed change history on the other.
Weight artifacts109 recognized artifacts · public access
03
License declarationapache-2.0Metadata declaration
04
Latest published source check
Repository SHA f68e6e470fd1…; a source check is not an inference benchmark or legal clearance. See the per-model evidence JSON for the exact source record.
A verification check does not pretend the model changed. Open Model Weights updates the verification date after a real primary-source check, while factual-update dates move only when retained source evidence changes.
Last verified
Latest primary-source check · full field verification
Last factual update
Latest retained field-level evidence change
Repository created
Source repository date; used only as a transparent release proxy when no stronger release date exists
Commercial useCommercial use allowed under apache-2.0 conditions
Repository license filesNone observed
Classification describes the checked license terms; it is not legal advice.
03 · HARDWARE
Weight-memory estimate
BF16 / FP16—
FP8 / INT8—
INT4—
Weight-only estimate; excludes runtime overhead.
04 · FORMATS
Formats & precision evidence
GGUFPyTorchSafetensors
BF16: not observedFP16: not observedFP8: not observedINT8: not observedINT4: not observedGGUF: observed
“Not observed” means no matching repository artifact/name/dtype signal was found; it does not prove no third-party conversion exists.
05 · LINEAGE
Base model & family
FamilyZONOS3
VariantMini
Declared base modelNot declared in model-card metadata
RelationNot declared
06 · TRAINING ASSETS
Recipe & data disclosure
Training recipenot-disclosed-in-checked-repository
Data disclosurenot-disclosed-in-standard-metadata-or-obvious-model-card-section
EvaluationsNot classified by automated verifier
Intermediate checkpointsNot classified by automated verifier
07 · RUNTIME SUPPORT
Declared, mentioned and externally evidenced runtimes
Declared libraryzonos3
Declared/taggedzonos3
Mentioned in model cardMLX, NeMo, SGLang
Artifact signalsNone
External primary runtime evidenceNone
Tested by Open Model WeightsNo — source-derived support only
Open Model Weights separates model-repository evidence from external runtime-release evidence. An external support statement does not change the official model repository's weight formats and does not imply an Open Model Weights inference benchmark.
Checked Hugging Face API metadata, exact repository file list, model card and config when accessible.
Field-by-field verification
Checked Hugging Face API metadata, exact repository file list, model card and config when accessible.
Field-by-field verification
Checked Hugging Face API metadata, exact repository file list, model card and config when accessible.
licenseverified
commercial useclassified
contextunknown
weight filesverified
formatsverified
base modelnot-declared
training recipenot-disclosed-in-checked-repository
data disclosurenot-disclosed-in-standard-metadata-or-obvious-model-card-section
runtime supportsource-derived-not-runtime-tested
MODELS FOR AI AGENTS · ROLE EVIDENCE
Where this model fits in an agent system
Specialized model. A supporting component. This checkpoint does not, by category alone, serve as the planning and tool-execution controller.
The curated notable tier is not an agent-readiness rating. Model-specific tool-call success, long-horizon task accuracy and compatible runtime configurations are not independently established here.
Source-linked license: apache-2.0 (allowed commercial-use classification; see original terms).