Skip to content
AI Atlas
ModelActiveOpen weightsIdentity probable

DeepSeek V3.2

DeepSeekfamily · DeepSeekhuggingface.co/deepseek-ai/DeepSeek-V3.2

DeepSeek-V3.2 is a large language model designed to harmonize high computational efficiency with strong reasoning and agentic tool-use performance. It introduces DeepSeek Sparse Attention (DSA), a fine-grained sparse attention mechanism...

Open in Graph
data quality72

Updated 5 h ago · first seen 11 Sept 2026

model_01M294WW8H2JQJXMSPY7QTDVWF

Overview

Identity

Canonical model
Yesidentity confidence: mediumOne 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
2
API aliases
deepseek-v3-2-reasoningdeepseek/deepseek-v3.2Identifiers 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 under MIT; commercial use allowed; redistribution allowed; derivatives allowed; 4 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

    Yes

  • Redistribution

    Yes

  • Derivatives

    Yes

Licence: MIT License (permissive · SPDX MIT · stated as “mit”)

dimensions marked null are unknown, not false

Key facts

Release date

Source:OpenRouter public model & pricing listingT2observed 15 h agomedium

Status

Source:DeepSeek — site & API docsT2observed 15 h agomediumLLM-extracted

Version

Source:DeepSeek — site & API docsT2observed 15 h agomediumLLM-extracted

Paper
huggingface.co/deepseek-ai/DeepSeek-V3.2/resolve/main/assets/paper.pdf

Source:DeepSeek — site & API docsT2observed 15 h agomediumLLM-extracted

Repository
huggingface.co/deepseek-ai/DeepSeek-V3.2

Source:DeepSeek — site & API docsT2observed 15 h agomediumLLM-extracted

Openrouter id

Source:OpenRouter public model & pricing listingT2observed 15 h agomedium

Architecture

Architecture

Source:Hugging Face Hub (public pages, model cards, papers)T2observed 15 h agomedium

Model type

Source:Hugging Face Hub (public pages, model cards, papers)T2observed 15 h agomedium

Parameters

Source:Hugging Face Hub (public pages, model cards, papers)T2observed 15 h agomedium

Tokenizer

Source:OpenRouter public model & pricing listingT2observed 15 h agomedium

Weights dtype

Source:Hugging Face Hub (public pages, model cards, papers)T2observed 15 h agomedium

File size

Source:Hugging Face Hub (public pages, model cards, papers)T2observed 15 h agomedium

Library name

Source:Hugging Face Hub (public pages, model cards, papers)T2observed 15 h agomedium

Pipeline tag

Source:Hugging Face Hub (public pages, model cards, papers)T2observed 15 h agomedium

Hugging Face repo

Source:OpenRouter public model & pricing listingT2observed 15 h agomedium

Capabilities

Modalities

Modalities
text
Input
text
Output
text

Capabilities

  • Tool calling

    Yes

    OpenRouter public model & pricing listing · T2

  • Structured output

    Yes

    OpenRouter public model & pricing listing · T2

  • Reasoning

    Yes

    OpenRouter public model & pricing listing · T2

  • Vision

    Unavailable

  • Audio

    Unavailable

  • Fine-tuning available

    Unavailable

Context window

Source:OpenRouter public model & pricing listingT2observed 12 h agomedium

Max output

Source:OpenRouter public model & pricing listingT2observed 15 h agomedium

Tokenizer

Source:OpenRouter public model & pricing listingT2observed 15 h agomedium

Benchmarks9

Compare with another model →

Comparable same task and conditions · Partially comparable same task, conditions differ (effort, temperature, judge) · Not comparable different variant or metric

Benchmark results grouped by comparability group
Benchmark · groupBest scoreTrustConfigurationResultsvs leaderEvaluatedSource
Terminal-Benchagentic · accuracy · variant=v2.1 · evaluator=Artificial AnalysisIndependentevaluatorArtificial Analysisvariantv2.11non-primary groupobs. 11 Sept 2026artificialanalysis.aiT2
Terminal-Benchagentic · accuracy · variant=hard · evaluator=Artificial AnalysisIndependentevaluatorArtificial Analysisvarianthard1−30.3 ptvs gpt-5.6-solobs. 11 Sept 2026artificialanalysis.aiT2
τ²-benchagentic · pass^1 · evaluator=Artificial AnalysisIndependentevaluatorArtificial Analysis1−8.48 ptvs Z.ai GLM 5.2obs. 11 Sept 2026artificialanalysis.aiT2
SWE-bench Multilingualcoding · resolved · board=Multilingual · system=mini-SWE-agentOfficial boardboardMultilingualsystemmini-SWE-agent1−13.7 ptvs gemini-3-flash13 Feb 2026swebench.comT2
SWE-bench Verifiedcoding · resolved · board=Verified · system=mini-SWE-agentOfficial boardboardVerifiedsystemmini-SWE-agentreasoning_efforthighconditions differ across rows → partially comparable1−6.80 ptvs Claude Opus 4.517 Feb 2026swebench.comT2
Artificial Analysis Intelligence Indexcomposite · indexIndependentversion4.31−31.9vs Claude Fable 5.1obs. 11 Sept 2026artificialanalysis.aiT2
IFBenchinstruction-following · accuracy · evaluator=Artificial AnalysisIndependentevaluatorArtificial Analysis1−22.6 ptvs Grok 4.3obs. 11 Sept 2026artificialanalysis.aiT2
Humanity's Last Examknowledge · accuracy · evaluator=Artificial AnalysisIndependentevaluatorArtificial Analysis1−34.6 ptvs Claude Fable 5.1obs. 11 Sept 2026artificialanalysis.aiT2
GPQA Diamondreasoning · accuracy · variant=GPQA Diamond · evaluator=Artificial AnalysisIndependentevaluatorArtificial AnalysisvariantGPQA Diamond1−12.2 ptvs gpt-6-astraobs. 11 Sept 2026artificialanalysis.aiT2

Current rows only, grouped by benchmark → canonical metric → comparability group (task configuration). Effort variants folded into this model appear as rows of the same group. 9 current rows in total. “vs leader” compares with the current leader of the benchmark's primary group only; other groups are not directly comparable. Comparability rules →

Providers & Pricing2

All offers in the price terminal →
Provider deployments, cheapest output first
ProviderContextInput / 1MCached inOutput / 1MStatusObservedSource
DeepSeek APIcheapest outputdeepseek/deepseek-v3.2163.8Kout 65.5K$0.135active6 h agosince 11 Sept 2026openrouter.aiT2
OpenRouterdeepseek/deepseek-v3.2163.8Kout 65.5K$0.135active2 h agosince 12 Sept 2026openrouter.aiT2

USD per 1M tokens as published by each provider; native units (per-request fees, flex/priority tiers) are kept verbatim. Rows are append-only — every price change is kept in the history below. Cost of a workload →

Price history

Step lines per provider; amber markers are recorded changes. Click a marker or a row for the evidence behind that price.

Output price · USD / 1M tokens 2 providers

Output price history of DeepSeek V3.2$0$0.10$0.20$0.30$0.40$0.50Sept 26Sept 26Sept 26Sept 26Sept 26DeepSeek API: first observed → $0.40 · 11 Sept 2026OpenRouter: first observed → $0.40 · 12 Sept 2026
  • DeepSeek API
  • OpenRouter
  • OpenRouterfirst observed $0.4012 Sept 2026
  • DeepSeek APIfirst observed $0.4011 Sept 2026

Input price · USD / 1M tokens 2 providers

Input price history of DeepSeek V3.2$0$0.10$0.20$0.30$0.40Sept 26Sept 26Sept 26Sept 26Sept 26DeepSeek API: first observed → $0.269 · 11 Sept 2026OpenRouter: first observed → $0.269 · 12 Sept 2026
  • DeepSeek API
  • OpenRouter
  • OpenRouterfirst observed $0.26912 Sept 2026
  • DeepSeek APIfirst observed $0.26911 Sept 2026

Hardware fit37

Estimated

3 of 37 device × quantization combinations fit.

Run locally: your machine →
Estimated hardware fit
HardwareQuantizationDevice memoryEst. memoryFits
Apple M3 Ultra4bit—394.6 GB est.Yes
Mac Studio (Apple M5 Ultra)4bit—394.6 GB est.Yes
NVIDIA DGX B2004bit1,440 GB394.6 GB est.Yes
Apple M2 Ultra4bit—394.6 GB est.No
Apple M1 Ultra4bit—394.6 GB est.No
Apple M3 Max4bit—394.6 GB est.No
Apple M4 Max4bit—394.6 GB est.No
Mac Studio (Apple M5 Max)4bit—394.6 GB est.No
MacBook Pro (Apple M5 Max)4bit—394.6 GB est.No
Apple M2 Max4bit—394.6 GB est.No
Apple M1 Max4bit—394.6 GB est.No
Apple M4 Pro4bit—394.6 GB est.No
Mac mini (Apple M5 Pro)4bit—394.6 GB est.No
MacBook Pro (Apple M5 Pro)4bit—394.6 GB est.No
Apple M3 Pro4bit—394.6 GB est.No
Apple M1 Pro4bit—394.6 GB est.No
Apple M2 Pro4bit—394.6 GB est.No
Apple M44bit—394.6 GB est.No
iMac (Apple M4)4bit—394.6 GB est.No
Mac mini (Apple M6)4bit—394.6 GB est.No
MacBook Air (Apple M5)4bit—394.6 GB est.No
MacBook Pro (Apple M5)4bit—394.6 GB est.No
Apple M24bit—394.6 GB est.No
Apple M34bit—394.6 GB est.No
Apple M14bit—394.6 GB est.No
NVIDIA GeForce RTX 30904bit24 GB394.6 GB est.No
NVIDIA GeForce RTX 40904bit24 GB394.6 GB est.No
NVIDIA GeForce RTX 50904bit32 GB394.6 GB est.No
NVIDIA A100 80GB4bit80 GB394.6 GB est.No
NVIDIA H100 SXM4bit80 GB394.6 GB est.No
NVIDIA H100 NVL4bit94 GB394.6 GB est.No
NVIDIA DGX Spark4bit128 GB394.6 GB est.No
NVIDIA H2004bit141 GB394.6 GB est.No
NVIDIA H200 NVL4bit141 GB394.6 GB est.No
NVIDIA B2004bit180 GB394.6 GB est.No
AMD Instinct MI300X4bit192 GB394.6 GB est.No
AMD Instinct MI325X4bit256 GB394.6 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 1DESCENDANTS 0 · ARTIFACTSDeepSeek-V3.2-Exp-BaseDeepSeekDeepSeek-V3.2-Exp-Base — DeepSeek2 quantizationsartifacts · collapsed2 quantizations — artifacts · collapsedDeepSeek V3.2685.4B params · this modelDeepSeek V3.2 — 685.4B params · this model

Versions & Artifacts0

Version history

Context window2 changes

11 Sept 2026→11 Sept 2026→12 Sept 2026current

Status1 change

11 Sept 2026→11 Sept 2026

Licensefirst observation only

11 Sept 2026current

Max outputfirst observation only

11 Sept 2026current

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.

Timeline2

Full timeline →
  • Listed by providerModelDeepSeek V3.2DeepSeek

    OpenRouter lists DeepSeek V3.2 at $0.269 in / $0.4 out per 1M tokens

    openrouter
    openrouter.ai
  • Context window changedModelDeepSeek V3.2DeepSeek

    DeepSeek V3.2: context length changed from 128000 to 163840

    Context window128K tokens→163.8K tokensopenrouter
    openrouter.ai

Change history

Temporal, append-only claims: a new observation closes the previous claim instead of overwriting it. Rewind the record with the as-of picker.
2 claims · 1 propertiesShow all properties

Statusstatus2

Claim history for Status
ValueValid from → toStatusSourceConfidenceExtractor
deprecated → supersededArtificial AnalysisT2mediumdeterministic
active → currentcurrentDeepSeek — site & API docsT2mediumllm

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

46

Source tiers

T246

Freshest observation

5 h ago

Conflicts

None

Source documents 9

Source documents
SourceDocumentTypeTierLast observedSnapshots
DeepSeek — site & API docsapi-docs.deepseek.com/news/news251201 newsT1· Official15 h ago1
Hugging Face Hub (public pages, model cards, papers)huggingface.co/deepseek-ai/DeepSeek-V3.2 model_pageT2· Quality secondary40 min ago5
Hugging Face Hub (public pages, model cards, papers)huggingface.co/cerebras/DeepSeek-V3.2-REAP-345B-A37B model_pageT2· Quality secondary41 min ago6
Hugging Face Hub (public pages, model cards, papers)huggingface.co/cerebras/DeepSeek-V3.2-REAP-508B-A37B model_pageT2· Quality secondary42 min ago6
Hugging Face Hub (public pages, model cards, papers)huggingface.co/models?author=deepseek-ai&p=0&sort=downloads listingT2· Quality secondary54 min ago8
OpenRouter public model & pricing listingopenrouter.ai/api/v1/models catalogueT2· Quality secondary2 h ago11
SWE-bench leaderboardsswebench.com/ leaderboardT2· Quality secondary5 h ago1
Artificial Analysisartificialanalysis.ai/leaderboards/models leaderboardT2· Quality secondary5 h ago2
Hugging Face Hub (public pages, model cards, papers)huggingface.co/deepseek-ai/DeepSeek-V3.2/raw/main/README.md model_cardT2· Quality secondary8 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 (72/100) measures how well AI Atlas knows this entity — completeness, primary-source ratio, freshness, conflicts — never how good the model is. Methodology →