Skip to content
ModelCap

Head-to-head comparison

DeepSeek V3.2 vs DeepSeek V4 Pro 0813

DeepSeek V3.2 (DeepSeek) and DeepSeek V4 Pro 0813 (DeepSeek) compared on the ModelCap Index, API price, context window, provider availability, weight access and every public benchmark board they share. Figures are the same ones shown on the live rankings; nothing here is a hidden score.

Snapshot as of 9 September 2026

As of 9 September 2026, DeepSeek V4 Pro 0813 holds the stronger ModelCap Index position (#20 vs #51); DeepSeek V3.2 is 4.3× cheaper per output token ($0.40/1M vs $1.74/1M); and DeepSeek V4 Pro 0813 offers the longer context window (1M vs 164K tokens).

Which should you choose?

Choose DeepSeek V3.2 if…

  • API cost matters — $0.40/1M output tokens against $1.74/1M, about 4.3× cheaper.
  • your workload is prompt-heavy — input tokens cost $0.269/1M against $0.579/1M.

Choose DeepSeek V4 Pro 0813 if…

  • you want the stronger overall ModelCap Index position — #20 against #51 (73.6 vs 58.2 points).
  • you need the longer context window — 1M tokens against 164K.
  • you want provider choice — 19 listed API providers against 15.
  • Arena coding is your yardstick — 1505 against 1476.
  • ARC-AGI-2 is your yardstick — 61.3% against 4.0%.
  • AA Terminal-Bench 2.1 is your yardstick — 78.7% against 46.8%.
  • AA Humanity's Last Exam is your yardstick — 41.0% against 11.2%.
  • you want the more recently listed model — DeepSeek V4 Pro 0813 was listed 12 August 2026, DeepSeek V3.2 1 December 2025.

DeepSeek V3.2 vs DeepSeek V4 Pro 0813: specs, pricing and context

Specification comparison of DeepSeek V3.2 and DeepSeek V4 Pro 0813
FieldDeepSeek V3.2DeepSeekDeepSeek V4 Pro 0813DeepSeek
ModelCap Index position#51#20
Index score58.273.6
EvidenceMeasuredMeasured
Input price / 1M tokens$0.269$0.579
Output price / 1M tokens$0.40$1.74
Context window164K tokens1M tokens
Max output tokens66K393K
API providers listed1519
Weight accessOpen weightsOpen weights
Input modalitiesTextText
First listed1 December 202512 August 2026
PublisherDeepSeekDeepSeek

Evidence: 4 public benchmark observations across 4 boards · 4 public benchmark observations across 4 boards. Prices are the lowest listed API offer per million tokens observed on the OpenRouter catalogue.

Benchmark scores: DeepSeek V3.2 vs DeepSeek V4 Pro 0813

Public benchmark boards where DeepSeek V3.2 or DeepSeek V4 Pro 0813 has a published result
BoardDeepSeek V3.2DeepSeek V4 Pro 0813Leads
Arena codingArena (LMArena)14761469–1483Thinking15051490–1519HighDeepSeek V4 Pro 0813
ARC-AGI-2ARC Prize Foundation4.0%61.3%MaxDeepSeek V4 Pro 0813
SWE-benchSWE-bench (Princeton / SWE-agent team)70.0%High—Only one result
AA Intelligence IndexArtificial Analysis—36.3intelligence-index:v4.3:maxOnly one result
AA Terminal-Bench 2.1Artificial Analysis46.8%terminal-bench:2.1:terminus-2:e2b:pass-at-1:3-repeats:source-model="DeepSeek V3.2":reasoning=true78.7%terminal-bench:2.1:terminus-2:e2b:pass-at-1:3-repeats:source-model="DeepSeek V4 Pro 0813 (max)":reasoning=trueDeepSeek V4 Pro 0813
AA τ²-Bench TelecomArtificial Analysis78.9%tau2:telecom:dual-control:pass-at-1:3-repeats:source-model="DeepSeek V3.2 (Non-reasoning)":reasoning=false—Only one result
AA Humanity's Last ExamArtificial Analysis11.2%hle:may-2025:text-only-2158:no-tools:pass-at-1:source-model="DeepSeek V3.2 (Non-reasoning)":reasoning=false41.0%hle:may-2025:text-only-2158:no-tools:pass-at-1:source-model="DeepSeek V4 Pro 0813 (max)":reasoning=trueDeepSeek V4 Pro 0813
WildClawBench OpenClawWildClawBench34.0—Only one result

Scores are the sources' own published figures for each model's best evaluated configuration; ModelCap never re-runs a benchmark.

Want a different pairing? Open the interactive comparison tool to swap either model for any current ranked language model.

DeepSeek V3.2 vs DeepSeek V4 Pro 0813: common questions

Is DeepSeek V3.2 better than DeepSeek V4 Pro 0813?

DeepSeek V4 Pro 0813 ranks higher on the ModelCap Index as of 9 September 2026: #20 against #51. That is a capability ranking built from public benchmark evidence with published uncertainty; whether it is "better" for you also depends on price, context and where you can run it. Their published uncertainty intervals do not overlap, but the aggregate Index does not predict performance on every task; compare the relevant benchmark configurations before choosing.

Is DeepSeek V3.2 cheaper than DeepSeek V4 Pro 0813?

DeepSeek V3.2 is cheaper on output tokens: $0.40/1M against $1.74/1M. Input tokens are $0.269/1M for DeepSeek V3.2 and $0.579/1M for DeepSeek V4 Pro 0813. Prices are the lowest listed API offer ModelCap observed, in USD per million tokens. For 1,000 requests with 2,000 input and 500 output tokens each (2M input + 0.5M output), the listed-rate estimate is $0.74 for DeepSeek V3.2 versus $2.03 for DeepSeek V4 Pro 0813. DeepSeek V3.2 costs 63.6% less in this scenario. This excludes caching, batch discounts, prompt-length tiers, tool charges and retries; verify the selected endpoint before budgeting.

Which has the bigger context window, DeepSeek V3.2 or DeepSeek V4 Pro 0813?

DeepSeek V4 Pro 0813 has the larger context window: 1M tokens against 164K.

Which is better for coding, DeepSeek V3.2 or DeepSeek V4 Pro 0813?

Arena coding: DeepSeek V3.2 1476, DeepSeek V4 Pro 0813 1505 — DeepSeek V4 Pro 0813 leads. AA Terminal-Bench 2.1: DeepSeek V3.2 46.8%, DeepSeek V4 Pro 0813 78.7% — DeepSeek V4 Pro 0813 leads.

Are DeepSeek V3.2 and DeepSeek V4 Pro 0813 open-weight models?

DeepSeek V3.2: Open weights. DeepSeek V4 Pro 0813: Open weights. Open weights mean the checkpoint can be downloaded and self-hosted under its licence; API-only models are available solely through hosted endpoints.

Where can I run DeepSeek V3.2 and DeepSeek V4 Pro 0813?

ModelCap currently lists 15 API providers for DeepSeek V3.2 and 19 for DeepSeek V4 Pro 0813, from the OpenRouter catalogue snapshot the site serves; each model page lists the providers and their prices.

How current is this DeepSeek V3.2 vs DeepSeek V4 Pro 0813 comparison?

Every figure comes from the sealed ModelCap dataset published 9 September 2026; the page re-renders within a minute of each data refresh, and the ModelCap Index positions are the same ones shown on the live rankings.

Explore further