Skip to content
AI Atlas
ModelActiveOpen weights

DeepSeek-V4-Flash-Base

DeepSeekfamily · DeepSeekhuggingface.co/deepseek-ai/DeepSeek-V4-Flash-Bas

Open in Graph
data quality54

Updated 5 h ago · first seen 11 Sept 2026

model_01M294Y7HXJW48RT7FTNJZA2TF

Overview

Identity

Canonical model
Yesidentity confidence: highOne row per real model release. Artifacts (checkpoints, quantisations, conversions) and folded evaluation variants point here.
Official checkpoints
official_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:Hugging Face Hub (public pages, model cards, papers)T2observed 14 h agomedium

Architecture

Architecture

Source:Hugging Face Hub (public pages, model cards, papers)T2observed 14 h agomedium

Model type

Source:Hugging Face Hub (public pages, model cards, papers)T2observed 14 h agomedium

Parameters

Source:Hugging Face Hub (public pages, model cards, papers)T2observed 14 h agomedium

Weights dtype

Source:Hugging Face Hub (public pages, model cards, papers)T2observed 14 h agomedium

File size

Source:Hugging Face Hub (public pages, model cards, papers)T2observed 14 h agomedium

Hugging Face repo

Source:Hugging Face Hub (public pages, model cards, papers)T2observed 14 h agomedium

Hardware fit37

Estimated

7 of 37 device × quantization combinations fit.

Run locally: your machine →
Estimated hardware fit
HardwareQuantizationDevice memoryEst. memoryFits
Apple M3 Ultra4bit—168.4 GB est.Yes
Mac Studio (Apple M5 Ultra)4bit—168.4 GB est.Yes
Apple M2 Ultra4bit—168.4 GB est.Yes
NVIDIA B2004bit180 GB168.4 GB est.Yes
AMD Instinct MI300X4bit192 GB168.4 GB est.Yes
AMD Instinct MI325X4bit256 GB168.4 GB est.Yes
NVIDIA DGX B2004bit1,440 GB168.4 GB est.Yes
Apple M1 Ultra4bit—168.4 GB est.No
Apple M3 Max4bit—168.4 GB est.No
Apple M4 Max4bit—168.4 GB est.No
Mac Studio (Apple M5 Max)4bit—168.4 GB est.No
MacBook Pro (Apple M5 Max)4bit—168.4 GB est.No
Apple M2 Max4bit—168.4 GB est.No
Apple M1 Max4bit—168.4 GB est.No
Apple M4 Pro4bit—168.4 GB est.No
Mac mini (Apple M5 Pro)4bit—168.4 GB est.No
MacBook Pro (Apple M5 Pro)4bit—168.4 GB est.No
Apple M3 Pro4bit—168.4 GB est.No
Apple M1 Pro4bit—168.4 GB est.No
Apple M2 Pro4bit—168.4 GB est.No
Apple M44bit—168.4 GB est.No
iMac (Apple M4)4bit—168.4 GB est.No
Mac mini (Apple M6)4bit—168.4 GB est.No
MacBook Air (Apple M5)4bit—168.4 GB est.No
MacBook Pro (Apple M5)4bit—168.4 GB est.No
Apple M24bit—168.4 GB est.No
Apple M34bit—168.4 GB est.No
Apple M14bit—168.4 GB est.No
NVIDIA GeForce RTX 30904bit24 GB168.4 GB est.No
NVIDIA GeForce RTX 40904bit24 GB168.4 GB est.No
NVIDIA GeForce RTX 50904bit32 GB168.4 GB est.No
NVIDIA A100 80GB4bit80 GB168.4 GB est.No
NVIDIA H100 SXM4bit80 GB168.4 GB est.No
NVIDIA H100 NVL4bit94 GB168.4 GB est.No
NVIDIA DGX Spark4bit128 GB168.4 GB est.No
NVIDIA H2004bit141 GB168.4 GB est.No
NVIDIA H200 NVL4bit141 GB168.4 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.

Versions & Artifacts0

Version history

Opennessfirst observation only

11 Sept 2026current

Parametersfirst 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 history18

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

Viewing AI Atlas as of 11 Sept 2026 — attributes exactly as the atlas knew them on that day; later corrections are not shown.

Back to today →

Attributes as of 11 Sept 2026 17 claims in force

Release date
22 Apr 2026
Openness
open-weights
Architecture
DeepseekV4ForCausalLM
Parameters
292B
Quantization format
fp8
File size
294.7 GB
Hugging Face repo
deepseek-ai/DeepSeek-V4-Flash-Base
Downloads
551,449
Likes
314
Gated
No
Is quantized
Yes
Last modified
2026-04-27T06:51:43+00:00
Downloads all time
1,108,069
Model type
deepseek_v4
Tags
safetensors, deepseek_v4, fp8
Weights dtype
BF16, F32, F8_E4M3, I64
18 claims · 18 properties

Release daterelease_date1

Claim history for Release date
ValueValid from → toStatusSourceConfidenceExtractor
22 Apr 2026 → currentcurrentHugging Face Hub (public pages, model cards, papers)T2mediumdeterministic

Opennessopenness1

Claim history for Openness
ValueValid from → toStatusSourceConfidenceExtractor
open-weights → currentcurrentHugging Face Hub (public pages, model cards, papers)T2mediumdeterministic

Architecturearchitecture1

Claim history for Architecture
ValueValid from → toStatusSourceConfidenceExtractor
DeepseekV4ForCausalLM → currentcurrentHugging Face Hub (public pages, model cards, papers)T2mediumdeterministic

Parametersparameter_count1

Claim history for Parameters
ValueValid from → toStatusSourceConfidenceExtractor
292B → currentcurrentHugging Face Hub (public pages, model cards, papers)T2mediumdeterministic

Quantization formatquant_format1

Claim history for Quantization format
ValueValid from → toStatusSourceConfidenceExtractor
fp8 → currentcurrentHugging Face Hub (public pages, model cards, papers)T2mediumdeterministic

File sizefile_size_gb1

Claim history for File size
ValueValid from → toStatusSourceConfidenceExtractor
294.7 GB → currentcurrentHugging Face Hub (public pages, model cards, papers)T2mediumdeterministic

Hugging Face repohf_repo1

Claim history for Hugging Face repo
ValueValid from → toStatusSourceConfidenceExtractor
deepseek-ai/DeepSeek-V4-Flash-Base → currentcurrentHugging Face Hub (public pages, model cards, papers)T2mediumdeterministic

Model cardmodel_card_url1

Claim history for Model card
ValueValid from → toStatusSourceConfidenceExtractor
https://huggingface.co/deepseek-ai/DeepSeek-V4-Flash-Base → currentcurrentHugging Face Hub (public pages, model cards, papers)T2mediumdeterministic

Downloadsmetric.downloads1

Claim history for Downloads
ValueValid from → toStatusSourceConfidenceExtractor
551,449 → currentcurrentHugging Face Hub (public pages, model cards, papers)T2mediumdeterministic

Likesmetric.likes1

Claim history for Likes
ValueValid from → toStatusSourceConfidenceExtractor
314 → currentcurrentHugging Face Hub (public pages, model cards, papers)T2mediumdeterministic

Gatedgated1

Claim history for Gated
ValueValid from → toStatusSourceConfidenceExtractor
No → currentcurrentHugging Face Hub (public pages, model cards, papers)T2mediumdeterministic

Is quantizedis_quantized1

Claim history for Is quantized
ValueValid from → toStatusSourceConfidenceExtractor
Yes → currentcurrentHugging Face Hub (public pages, model cards, papers)T2mediumdeterministic

Last modifiedlast_modified1

Claim history for Last modified
ValueValid from → toStatusSourceConfidenceExtractor
2026-04-27T06:51:43+00:00 → currentcurrentHugging Face Hub (public pages, model cards, papers)T2mediumdeterministic

Downloads all timemetric.downloads_all_time1

Claim history for Downloads all time
ValueValid from → toStatusSourceConfidenceExtractor
1,108,069 → currentcurrentHugging Face Hub (public pages, model cards, papers)T2mediumdeterministic

Model typemodel_type1

Claim history for Model type
ValueValid from → toStatusSourceConfidenceExtractor
deepseek_v4 → currentcurrentHugging Face Hub (public pages, model cards, papers)T2mediumdeterministic

Tagstags1

Claim history for Tags
ValueValid from → toStatusSourceConfidenceExtractor
safetensors, deepseek_v4, fp8 → currentcurrentHugging Face Hub (public pages, model cards, papers)T2mediumdeterministic

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

Weights dtypeweights_dtype1

Claim history for Weights dtype
ValueValid from → toStatusSourceConfidenceExtractor
BF16, F32, F8_E4M3, I64 → currentcurrentHugging Face Hub (public pages, model cards, papers)T2mediumdeterministic

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

19

Source tiers

T219

Freshest observation

5 h ago

Conflicts

None

Source documents 2

Source documents
SourceDocumentTypeTierLast observedSnapshots
Hugging Face Hub (public pages, model cards, papers)huggingface.co/models?author=deepseek-ai&p=0&sort=downloads listingT2· Quality secondary2 min ago8
Hugging Face Hub (public pages, model cards, papers)huggingface.co/deepseek-ai/DeepSeek-V4-Flash-Base model_pageT2· Quality secondary8 h ago4

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

Data quality (54/100) measures how well AI Atlas knows this entity — completeness, primary-source ratio, freshness, conflicts — never how good the model is. Methodology →