Skip to content
AI Atlas
ModelPreviewOpen weights

QVQ-72B-Preview

Qwen Teamfamily · Qwen

Open in Graph
data quality56

Updated 3 h ago · first seen 12 Sept 2026

model_01M29X5PPPHD00X3M8KZ1VYV79

Overview

Identity

Canonical model
Yesidentity confidence: highOne row per real model release. Artifacts (checkpoints, quantisations, conversions) and folded evaluation variants point here.
Official checkpoints
None recordedofficial_checkpoints = hf_repo identifiers carried by the model itself; artifacts are separate entities pointing here through canonical_id.
Artifacts
None recordedSeparate entities (checkpoint · quantization · conversion · packaging) pointing to this model through canonical_id.
Provider deployments
None recorded
API aliases
NoneIdentifiers under which providers and evaluators refer to this model.
Folded evaluation variants
0Effort / thinking variants (…-high, …-non-reasoning) are result configurations of this model, not separate models. Their old URLs redirect here.

Openness

Open weights— weights downloadable; 7 dimensions unknown.

Weights downloadable under a permissive or Creative Commons licence allowing commercial use; code or data may be missing.

  • Weights

    Yes

  • Inference code

    —

  • Training code

    —

  • Training data

    —

  • Dataset

    —

  • Commercial use

    —

  • Redistribution

    —

  • Derivatives

    —

dimensions marked null are unknown, not false

Key facts

Release date

Source:Qwen — official blogT2observed 13 h agomediumLLM-extracted

Status

Source:Qwen — official blogT2observed 13 h agomediumLLM-extracted

Version

Source:Qwen — official blogT2observed 13 h agomediumLLM-extracted

Architecture

Architecture

Source:Qwen — official blogT2observed 13 h agomediumLLM-extracted

Parameters

Source:Qwen — official blogT2observed 13 h agomediumLLM-extracted

Capabilities

Modalities

Modalities
imagetext
Input
imagetext
Output
text

Capabilities

  • Tool calling

    Unavailable

  • Structured output

    Unavailable

  • Reasoning

    Yes

    Qwen — official blog · T2

  • Vision

    Yes

    Qwen — official blog · T2

  • Audio

    Unavailable

  • Fine-tuning available

    Unavailable

No structured attributes yet.

Hardware fit37

Estimated

23 of 37 device × quantization combinations fit.

Run locally: your machine →
Estimated hardware fit
HardwareQuantizationDevice memoryEst. memoryFits
Apple M3 Ultra4bit—41.9 GB est.Yes
Mac Studio (Apple M5 Ultra)4bit—41.9 GB est.Yes
Apple M2 Ultra4bit—41.9 GB est.Yes
Apple M1 Ultra4bit—41.9 GB est.Yes
Apple M3 Max4bit—41.9 GB est.Yes
Apple M4 Max4bit—41.9 GB est.Yes
Mac Studio (Apple M5 Max)4bit—41.9 GB est.Yes
MacBook Pro (Apple M5 Max)4bit—41.9 GB est.Yes
Apple M2 Max4bit—41.9 GB est.Yes
Apple M1 Max4bit—41.9 GB est.Yes
Apple M4 Pro4bit—41.9 GB est.Yes
Mac mini (Apple M5 Pro)4bit—41.9 GB est.Yes
MacBook Pro (Apple M5 Pro)4bit—41.9 GB est.Yes
NVIDIA A100 80GB4bit80 GB41.9 GB est.Yes
NVIDIA H100 SXM4bit80 GB41.9 GB est.Yes
NVIDIA H100 NVL4bit94 GB41.9 GB est.Yes
NVIDIA DGX Spark4bit128 GB41.9 GB est.Yes
NVIDIA H2004bit141 GB41.9 GB est.Yes
NVIDIA H200 NVL4bit141 GB41.9 GB est.Yes
NVIDIA B2004bit180 GB41.9 GB est.Yes
AMD Instinct MI300X4bit192 GB41.9 GB est.Yes
AMD Instinct MI325X4bit256 GB41.9 GB est.Yes
NVIDIA DGX B2004bit1,440 GB41.9 GB est.Yes
Apple M3 Pro4bit—41.9 GB est.No
Apple M1 Pro4bit—41.9 GB est.No
Apple M2 Pro4bit—41.9 GB est.No
Apple M44bit—41.9 GB est.No
iMac (Apple M4)4bit—41.9 GB est.No
Mac mini (Apple M6)4bit—41.9 GB est.No
MacBook Air (Apple M5)4bit—41.9 GB est.No
MacBook Pro (Apple M5)4bit—41.9 GB est.No
Apple M24bit—41.9 GB est.No
Apple M34bit—41.9 GB est.No
Apple M14bit—41.9 GB est.No
NVIDIA GeForce RTX 30904bit24 GB41.9 GB est.No
NVIDIA GeForce RTX 40904bit24 GB41.9 GB est.No
NVIDIA GeForce RTX 50904bit32 GB41.9 GB est.No
Assumptions (7)
  • Estimated, not measured: weights = parameters × bytes/param × 1.15 runtime overhead (or the observed artifact file size when one is recorded).
  • bytes/param: 4bit = 0.5, 8bit = 1.0, fp16 = 2.0 (uniform quantization, no per-layer exceptions).
  • KV cache: 2 × layers × kv_heads × head_dim × 2 bytes × context × batch when the architecture is known; otherwise 0.5 GB per 8 192 tokens (× batch), independent of architecture (GQA/MLA models need less).
  • A model 'fits' when the estimate is at most the device memory minus 2 GB reserved for the OS and framework.
  • Mixture-of-experts models are estimated on total parameters (all experts must be resident); active parameters are ignored.
  • Device memory uses the largest configuration when several are listed (e.g. Apple silicon tiers).
  • Multi-GPU: device memories are summed; interconnect bandwidth, tensor-parallel replication and pipeline bubbles are not modelled.

Lineage

Open in Graph →
Explicit derived_from / fine_tuned_from / distilled_from relations stated by sources; artifacts collapsed by kind.
ANCESTORS 1Qwen2-VL-72BQwen TeamQwen2-VL-72B — Qwen TeamQVQ-72B-Preview72B params · this modelQVQ-72B-Preview — 72B params · this model

Versions & Artifacts0

Version history

Opennessfirst observation only

11 Sept 2026current

Parametersfirst observation only

11 Sept 2026current

Statusfirst observation only

11 Sept 2026current

Each hop is a claim: click a value for its source, tier and observation time. Nothing is overwritten — a new observation closes the previous claim.

Artifacts 0

No artifact (checkpoint, quantisation, conversion or packaging) points to this model yet.

Change history14

Temporal, append-only claims: a new observation closes the previous claim instead of overwriting it. Rewind the record with the as-of picker.
14 claims · 14 properties

Familyfamily1

Claim history for Family
ValueValid from → toStatusSourceConfidenceExtractor
Qwen → currentcurrentQwen — official blogT2mediumllm

Versionversion1

Claim history for Version
ValueValid from → toStatusSourceConfidenceExtractor
72B-Preview → currentcurrentQwen — official blogT2mediumllm

Release daterelease_date1

Claim history for Release date
ValueValid from → toStatusSourceConfidenceExtractor
25 Dec 2024 → currentcurrentQwen — official blogT2mediumllm

Statusstatus1

Claim history for Status
ValueValid from → toStatusSourceConfidenceExtractor
preview → currentcurrentQwen — official blogT2mediumllm

Opennessopenness1

Claim history for Openness
ValueValid from → toStatusSourceConfidenceExtractor
open-weights → currentcurrentQwen — official blogT2mediumllm

Architecturearchitecture1

Claim history for Architecture
ValueValid from → toStatusSourceConfidenceExtractor
multimodal reasoning → currentcurrentQwen — official blogT2mediumllm

Parametersparameter_count1

Claim history for Parameters
ValueValid from → toStatusSourceConfidenceExtractor
72B → currentcurrentQwen — official blogT2mediumllm

Modalitiesmodalities1

Claim history for Modalities
ValueValid from → toStatusSourceConfidenceExtractor
image, text → currentcurrentQwen — official blogT2mediumllm

Input modalitiesmodalities_input1

Claim history for Input modalities
ValueValid from → toStatusSourceConfidenceExtractor
image, text → currentcurrentQwen — official blogT2mediumllm

Output modalitiesmodalities_output1

Claim history for Output modalities
ValueValid from → toStatusSourceConfidenceExtractor
text → currentcurrentQwen — official blogT2mediumllm

Reasoningreasoning1

Claim history for Reasoning
ValueValid from → toStatusSourceConfidenceExtractor
Yes → currentcurrentQwen — official blogT2mediumllm

Safety notessafety_notes1

Claim history for Safety notes
ValueValid from → toStatusSourceConfidenceExtractor
The model requires enhanced safety measures to ensure reliable and secure performance. Limitations include language mixing, recursive reasoning (circular logic), and potential hallucinations due to losing focus on image content during multi-step reasoning. → currentcurrentQwen — official blogT2mediumllm

Visionvision1

Claim history for Vision
ValueValid from → toStatusSourceConfidenceExtractor
Yes → currentcurrentQwen — official blogT2mediumllm

Weights availableweights_available1

Claim history for Weights available
ValueValid from → toStatusSourceConfidenceExtractor
Yes → currentcurrentAI Atlas curated registry (YAML, versioned in git, every entry carries its source URL)T2highderived

Claims are temporal and append-only: a new observation closes the previous claim (valid_to) instead of overwriting it. Conflicting claims from different sources are kept side by side and flagged — never averaged. Methodology →

Provenance

Attributed facts

14

Source tiers

T214

Freshest observation

3 h ago

Conflicts

None

Source documents 2

Source documents
SourceDocumentTypeTierLast observedSnapshots
Qwen — official blogqwenlm.github.io/blog/qvq-72b-preview newsT1· Official13 h ago1
Qwen — official blogqwenlm.github.io/blog/qvq-max-preview newsT1· Official13 h ago1

Tier 1 = official/primary, 2 = quality secondary, 3 = community, 4 = unverified. Every snapshot is archived; see all sources and the methodology.

Data quality (56/100) measures how well AI Atlas knows this entity — completeness, primary-source ratio, freshness, conflicts — never how good the model is. Methodology →