Why is this comparison indexable?

It has 6 verifiable comparison points, and both models have pricing or benchmark data.

Are missing metrics invented?

No. Metrics without LMSpeed data are omitted from this report.

Back to models

Data points: 142

Key Takeaways

The readout for DeepSeek V3.2 and DeepSeek V4 Flash, before the detailed comparison sheet.

Decision read

DeepSeek V4 Flash

DeepSeek V4 Flash currently has the stronger profile, with verified wins split 1 to 5.

Evidence depth

142 data points

Includes 28 benchmark rows, 4 audit samples, and 8 provider examples.

Selection signal

Start with DeepSeek V4 Flash

The charts below split 40 high-signal samples across speed, scores, and audit health.

Change comparison

Switch either side of this report to compare another model with the same LMSpeed data pipeline.

Model AModel B

Select a different model to open a new comparison URL.

Model compare DeepSeek V3.2 vs DeepSeek V4 Flash deepseek-v3-2-vs-deepseek-v4-flash	Model A DeepSeek V3.2	Model B DeepSeek V4 Flash
Overall leader	Contender	Leading
Verified metric wins	1 wins	5 wins
Where it leads	First-token latency	Cheapest input price, Average speed, Free providers, Provider coverage, Recent tests
Model metadata	DeepSeek V3.2 exposes 131.1K tokens; notable signals: Text input, Text output, Tool calling, Structured outputs.	DeepSeek V4 Flash exposes 1.0M tokens; notable signals: Text input, Text output, Tool calling, Structured outputs.
Developer	DeepSeek	DeepSeek
Context window	131.1K tokens	1.0M tokens
Max output	64K tokens	16.4K tokens
Released	Dec 2025	Apr 2026
Modalities	Input Text Output Text	Input Text Output Text
Features	Text inputText outputTool callingStructured outputsJSON modeReasoning	Text inputText outputTool callingStructured outputsJSON modeReasoning
Parameters	No data	No data
Tokenizer	DeepSeek	DeepSeek
Knowledge cutoff	No data	No data
OpenRouter ID	deepseek/deepseek-v3.2	deepseek/deepseek-v4-flash
References	No data	No data

Benchmark score comparison

Third-party benchmark profile synced into LMSpeed; only metrics available for both models are shown.

Category performance

Compare benchmark category scores on a 0-100 scale. Select a category to inspect the gap.

Model A coverage: 6 / 8
Model B coverage: 7 / 8
Shared: 6 shared categories

DeepSeek V3.2DeepSeek V4 Flash

Avg. score

DeepSeek V3.2

54.3

Avg. score

DeepSeek V4 Flash

Selected category

Instruction following

DeepSeek V4 Flash leads by 24.5

Professional benchmark details

Metric-level scores with benchmark source, rank depth, confidence, error, and evaluation date where available.

Group

Aggregateverified

BenchLM overall score

SourceDeepSeek V3.2

DeepSeek V3.2

56.0

Rank #38/82 · confidence 1 · eval date 2025-12-01

+14.0

DeepSeek V4 Flash

winner

70.0

Rank #23/82 · confidence 2 · eval date 2026-04-24

Pricingverified

Output price

SourceDeepSeek V3.2 (Non-reasoning)

DeepSeek V3.2

$0.420/M

Rank #12/95 · confidence 4

+$0.140/M

DeepSeek V4 Flash

winner

$0.280/M

Rank #2/95 · confidence 4

Pricingverified

Blended price

SourceDeepSeek V3.2 (Non-reasoning)

DeepSeek V3.2

$0.315/M

Rank #19/95 · confidence 4

+$0.140/M

DeepSeek V4 Flash

winner

$0.175/M

Rank #6/95 · confidence 4

Pricingverified

Input price

SourceDeepSeek V3.2 (Non-reasoning)

DeepSeek V3.2

$0.280/M

Rank #25/95 · confidence 4

+$0.140/M

DeepSeek V4 Flash

winner

$0.140/M

Rank #10/95 · confidence 4

Reasoningverified

MMLU-Pro

SourceDeepSeek V3.2 (Non-reasoning)

DeepSeek V3.2

83.7%

Rank #24/74 · confidence 4

+2.7%

DeepSeek V4 Flash

winner

86.4%

Rank #12/74 · confidence 2 · eval date 2026-04-24

Reasoningverified

HLE

SourceDeepSeek V3.2 (Non-reasoning)

DeepSeek V3.2

10.5%

Rank #54/103 · confidence 4

+17.3%

DeepSeek V4 Flash

winner

27.8%

Rank #20/103 · confidence 4

Reasoningverified

CritPt

SourceDeepSeek V3.2

DeepSeek V3.2

0.9

Rank #36/44 · confidence 1 · eval date 2025-12-01

+2.5

DeepSeek V4 Flash

winner

3.4

Rank #21/44 · confidence 2 · eval date 2026-04-24

Reasoningverified

GPQA

SourceDeepSeek V3.2 (Non-reasoning)

DeepSeek V3.2

75.1%

Rank #57/103 · confidence 4

+11.6%

DeepSeek V4 Flash

winner

86.7%

Rank #22/103 · confidence 4

Reasoningverified

AA-LCR

SourceDeepSeek V3.2

DeepSeek V3.2

39.0

Rank #49/57 · confidence 1 · eval date 2025-12-01

+23.7

DeepSeek V4 Flash

winner

62.7

Rank #29/57 · confidence 2 · eval date 2026-04-24

Codingverified

LiveCodeBench

SourceDeepSeek V3.2 (Non-reasoning)

DeepSeek V3.2

59.3%

Rank #32/70 · confidence 4

+29.1%

DeepSeek V4 Flash

winner

88.4%

Rank #4/70 · confidence 2 · eval date 2026-04-24

Codingverified

BenchLM Coding score

SourceDeepSeek V3.2

DeepSeek V3.2

56.8

Rank #37/69 · confidence 1 · eval date 2025-12-01

+24.8

DeepSeek V4 Flash

winner

81.6

Rank #14/69 · confidence 2 · eval date 2026-04-24

Codingverified

Terminal-Bench Hard

SourceDeepSeek V3.2

DeepSeek V3.2

32.6

Rank #32/59 · confidence 1 · eval date 2025-12-01

+6.0

DeepSeek V4 Flash

winner

38.6

Rank #23/59 · confidence 2 · eval date 2026-04-24

Codingverified

AA Coding Index

SourceDeepSeek V3.2

DeepSeek V3.2

34.6

Rank #36/61 · confidence 1 · eval date 2025-12-01

+5.2

DeepSeek V4 Flash

winner

39.8

Rank #24/61 · confidence 2 · eval date 2026-04-24

Codingverified

AA-SciCode

SourceDeepSeek V3.2

DeepSeek V3.2

38.7

Rank #41/63 · confidence 1 · eval date 2025-12-01

+3.3

DeepSeek V4 Flash

winner

42.0

Rank #30/63 · confidence 2 · eval date 2026-04-24

Codingverified

SciCode

SourceDeepSeek V3.2 (Non-reasoning)

DeepSeek V3.2

38.7%

Rank #53/101 · confidence 4

+3.3%

DeepSeek V4 Flash

winner

42.0%

Rank #37/101 · confidence 4

Knowledgeverified

AA-Omniscience Hallucination Rate

SourceDeepSeek V3.2

DeepSeek V3.2

winner

93.5

Rank #1/58 · confidence 1 · eval date 2025-12-01

+3.8

DeepSeek V4 Flash

89.7

Rank #9/58 · confidence 2 · eval date 2026-04-24

Knowledgeverified

AA-HLE

SourceDeepSeek V3.2

DeepSeek V3.2

10.5

Rank #41/63 · confidence 1 · eval date 2025-12-01

+17.3

DeepSeek V4 Flash

winner

27.8

Rank #17/63 · confidence 2 · eval date 2026-04-24

Knowledgeverified

AA-GPQA Diamond

SourceDeepSeek V3.2

DeepSeek V3.2

75.1

Rank #43/63 · confidence 1 · eval date 2025-12-01

+11.6

DeepSeek V4 Flash

winner

86.7

Rank #19/63 · confidence 2 · eval date 2026-04-24

Knowledgeverified

AA-Omniscience Accuracy

SourceDeepSeek V3.2

DeepSeek V3.2

24.2

Rank #41/58 · confidence 1 · eval date 2025-12-01

+11.3

DeepSeek V4 Flash

winner

35.5

Rank #23/58 · confidence 2 · eval date 2026-04-24

Knowledgeverified

Artificial Analysis Intelligence Index

SourceDeepSeek V3.2

DeepSeek V3.2

24.7

Rank #41/66 · confidence 1 · eval date 2025-12-01

+12.7

DeepSeek V4 Flash

winner

37.4

Rank #24/66 · confidence 2 · eval date 2026-04-24

Knowledgeverified

BenchLM Knowledge score

SourceDeepSeek V3.2

DeepSeek V3.2

61.8

Rank #33/67 · confidence 1 · eval date 2025-12-01

0.0

DeepSeek V4 Flash

61.8

Rank #33/67 · confidence 2 · eval date 2026-04-24

Multimodalverified

Design Arena Website

SourceDeepSeek V3.2

DeepSeek V3.2

1218.0

Rank #27/46 · confidence 1 · eval date 2025-12-01

+36.0

DeepSeek V4 Flash

winner

1254.0

Rank #20/46 · confidence 2 · eval date 2026-04-24

Instruction followingverified

AA-IFBench

SourceDeepSeek V3.2

DeepSeek V3.2

49.0

Rank #35/60 · confidence 1 · eval date 2025-12-01

+24.5

DeepSeek V4 Flash

winner

73.5

Rank #12/60 · confidence 2 · eval date 2026-04-24

Agentsverified

Tau2-Telecom

SourceDeepSeek V3.2

DeepSeek V3.2

78.9

Rank #34/58 · confidence 1 · eval date 2025-12-01

+16.7

DeepSeek V4 Flash

winner

95.6

Rank #9/58 · confidence 2 · eval date 2026-04-24

Agentsverified

Claw-Eval

SourceDeepSeek V3.2

DeepSeek V3.2

40.2

Rank #21/21 · confidence 1 · eval date 2025-12-01

+17.6

DeepSeek V4 Flash

winner

57.8

Rank #10/21 · confidence 2 · eval date 2026-04-24

Agentsverified

AA Agentic Index

SourceDeepSeek V3.2

DeepSeek V3.2

39.8

Rank #36/57 · confidence 1 · eval date 2025-12-01

+22.5

DeepSeek V4 Flash

winner

62.3

Rank #12/57 · confidence 2 · eval date 2026-04-24

Agentsverified

Gert Labs

SourceDeepSeek V3.2

DeepSeek V3.2

29.6

Rank #35/36 · confidence 1 · eval date 2025-12-01

+24.8

DeepSeek V4 Flash

winner

54.4

Rank #13/36 · confidence 2 · eval date 2026-04-24

Agentsverified

BenchLM Agentic score

SourceDeepSeek V3.2

DeepSeek V3.2

49.7

Rank #38/69 · confidence 1 · eval date 2025-12-01

+10.9

DeepSeek V4 Flash

winner

60.6

Rank #24/69 · confidence 2 · eval date 2026-04-24

API audit comparison

Latest completed audits from shared providers, with four safety and integrity score groups plus report links.

Provider	DeepSeek V3.2	DeepSeek V4 Flash
PICO APIWinner: DeepSeek V4 Flash	DeepSeek V3.2 deepseek-v3.2 No audit yet	DeepSeek V4 Flash deepseek-v4-flash Audit score 92 View report 10068100100
速创APIWinner: DeepSeek V4 Flash	DeepSeek V3.2 deepseek-v3.2 No audit yet	DeepSeek V4 Flash deepseek-v4-flash Audit score 92 View report 10068100100
CatClaw APIWinner: DeepSeek V4 Flash	DeepSeek V3.2 deepseek-v3.2 No audit yet	DeepSeek V4 Flash deepseek-v4-flash Audit score 83 View report 766886100
小水管 APIWinner: DeepSeek V4 Flash	DeepSeek V3.2 deepseek-v3.2 No audit yet	DeepSeek V4 Flash deepseek-v4-flash Audit score 83 View report 807280100

Provider examples

Speed aggregates and input/output pricing share each provider row for real API selection and migration cost checks.

Provider	DeepSeek V3.2	DeepSeek V4 Flash
SiliconFlow151 tests	DeepSeek V3.2 speed / latency 26 tok/s / 11125ms input / output No data	DeepSeek V4 Flash speed / latency 45 tok/s / 9731ms input / output No data
云智API80 tests	DeepSeek V3.2 speed / latency 29 tok/s / 3036ms input / output No data	DeepSeek V4 Flash speed / latency N/A / N/A input / output No data
阿里云百炼 DashScope75 tests	DeepSeek V3.2 speed / latency 27 tok/s / 1295ms input / output No data	DeepSeek V4 Flash speed / latency 108 tok/s / 3324ms input / output No data
NVIDIA NIM60 tests	DeepSeek V3.2 speed / latency 18 tok/s / 4488ms input / output No data	DeepSeek V4 Flash speed / latency 23 tok/s / 14357ms input / output No data
火山引擎 Ark50 tests	DeepSeek V3.2 speed / latency 29 tok/s / 1828ms input / output No data	DeepSeek V4 Flash speed / latency 73 tok/s / 3161ms input / output No data
MapleLeaf API	DeepSeek V3.2 deepseek-ai/deepseek-v3.2 speed / latency No data input / output $0/M / $0/M	DeepSeek V4 Flash deepseek-v4-flash-nothinking speed / latency No data input / output $75.00/M / $75.00/M
MineWuer API	DeepSeek V3.2 linuxdo/deepseek-v3.2-chat speed / latency No data input / output $0/M / $0/M	DeepSeek V4 Flash deepseek-v4-flash speed / latency No data input / output $0/M / $0/M
wuer的api站	DeepSeek V3.2 deepseek-ai/deepseek-v3.2 speed / latency No data input / output $0/M / $0/M	DeepSeek V4 Flash deepseek-v4-flash speed / latency No data input / output $0/M / $0/M

When to choose each model

This report only uses LMSpeed data for DeepSeek V3.2 and DeepSeek V4 Flash: pricing, speed aggregates, third-party benchmark scores, and shared provider samples.

Guidance	DeepSeek V3.2	DeepSeek V4 Flash
When to choose each model	DeepSeek V3.2 DeepSeek V3.2 is stronger when you prioritize First-token latency.	DeepSeek V4 Flash DeepSeek V4 Flash is stronger when you prioritize Cheapest input price, Average speed, Free providers, Provider coverage, Recent tests.

FAQ

TL;DR: DeepSeek V4 Flash leads across 142 verifiable data points, including pricing, speed, latency, benchmarks, and provider examples.

Why is this comparison indexable?: It has 6 verifiable comparison points, and both models have pricing or benchmark data.
Are missing metrics invented?: No. Metrics without LMSpeed data are omitted from this report.

Continue from DeepSeek V3.2 vs DeepSeek V4 Flash into nearby model comparisons with enough verified LMSpeed data.

DeepSeek V3.2 vs GPT 5 4deepseek-v3-2-vs-gpt-5-4 Claude Opus 4 6 vs DeepSeek V3.2claude-opus-4-6-vs-deepseek-v3-2 DeepSeek V3.2 vs GPT 5deepseek-v3-2-vs-gpt-5 DeepSeek V3.2 vs GPT 5 2deepseek-v3-2-vs-gpt-5-2

Data as of Jul 3, 2026, 02:14 PMRankings are based on community-submitted tests and periodic health probes. Advisory only, not official data.

Model compare

DeepSeek V3.2 vs DeepSeek V4 Flash

deepseek-v3-2-vs-deepseek-v4-flash

Model A

DeepSeek V3.2

Model B

DeepSeek V4 Flash

Overall leader

Contender

Leading

Verified metric wins

1 wins

5 wins

Where it leads

First-token latency

Cheapest input price, Average speed, Free providers, Provider coverage, Recent tests

Model metadata

DeepSeek V3.2 exposes 131.1K tokens; notable signals: Text input, Text output, Tool calling, Structured outputs.

DeepSeek V4 Flash exposes 1.0M tokens; notable signals: Text input, Text output, Tool calling, Structured outputs.

Developer

DeepSeek

Context window

131.1K tokens

1.0M tokens

Max output

64K tokens

16.4K tokens

Released

Dec 2025

Apr 2026

Modalities

Input

Text

Output

Text

Input

Text

Output

Text

Features

Text inputText outputTool callingStructured outputsJSON modeReasoning

Parameters

No data

Tokenizer

DeepSeek

Knowledge cutoff

No data

OpenRouter ID

deepseek/deepseek-v3.2

deepseek/deepseek-v4-flash

References

No data

Provider

DeepSeek V3.2

DeepSeek V4 Flash

PICO APIWinner: DeepSeek V4 Flash

DeepSeek V3.2

deepseek-v3.2

No audit yet

DeepSeek V4 Flash

deepseek-v4-flash

Audit score

View report

10068100100

速创APIWinner: DeepSeek V4 Flash

DeepSeek V3.2

deepseek-v3.2

No audit yet

DeepSeek V4 Flash

deepseek-v4-flash

Audit score

View report

10068100100

CatClaw APIWinner: DeepSeek V4 Flash

DeepSeek V3.2

deepseek-v3.2

No audit yet

DeepSeek V4 Flash

deepseek-v4-flash

Audit score

View report

766886100

小水管 APIWinner: DeepSeek V4 Flash

DeepSeek V3.2

deepseek-v3.2

No audit yet

DeepSeek V4 Flash

deepseek-v4-flash

Audit score

View report

807280100

Provider

DeepSeek V3.2

DeepSeek V4 Flash

SiliconFlow151 tests

DeepSeek V3.2

speed / latency

26 tok/s / 11125ms

input / output

No data

DeepSeek V4 Flash

speed / latency

45 tok/s / 9731ms

input / output

No data

云智API80 tests

DeepSeek V3.2

speed / latency

29 tok/s / 3036ms

input / output

No data

DeepSeek V4 Flash

speed / latency

N/A / N/A

input / output

No data

阿里云百炼 DashScope75 tests

DeepSeek V3.2

speed / latency

27 tok/s / 1295ms

input / output

No data

DeepSeek V4 Flash

speed / latency

108 tok/s / 3324ms

input / output

No data

NVIDIA NIM60 tests

DeepSeek V3.2

speed / latency

18 tok/s / 4488ms

input / output

No data

DeepSeek V4 Flash

speed / latency

23 tok/s / 14357ms

input / output

No data

火山引擎 Ark50 tests

DeepSeek V3.2

speed / latency

29 tok/s / 1828ms

input / output

No data

DeepSeek V4 Flash

speed / latency

73 tok/s / 3161ms

input / output

No data

MapleLeaf API

DeepSeek V3.2

deepseek-ai/deepseek-v3.2

speed / latency

No data

input / output

$0/M / $0/M

DeepSeek V4 Flash

deepseek-v4-flash-nothinking

speed / latency

No data

input / output

$75.00/M / $75.00/M

MineWuer API

DeepSeek V3.2

linuxdo/deepseek-v3.2-chat

speed / latency

No data

input / output

$0/M / $0/M

DeepSeek V4 Flash

deepseek-v4-flash

speed / latency

No data

input / output

$0/M / $0/M

wuer的api站

DeepSeek V3.2

deepseek-ai/deepseek-v3.2

speed / latency

No data

input / output

$0/M / $0/M

DeepSeek V4 Flash

deepseek-v4-flash

speed / latency

No data

input / output

$0/M / $0/M

Guidance

DeepSeek V3.2

DeepSeek V4 Flash

When to choose each model

DeepSeek V3.2

DeepSeek V3.2 is stronger when you prioritize First-token latency.

DeepSeek V4 Flash

DeepSeek V4 Flash is stronger when you prioritize Cheapest input price, Average speed, Free providers, Provider coverage, Recent tests.

Key Takeaways

Change comparison

DeepSeek V3.2 vs DeepSeek V4 Flash

DeepSeek V3.2

DeepSeek V4 Flash

Benchmark score comparison

Category performance

Professional benchmark details

BenchLM overall score

Output price

Blended price

Input price

MMLU-Pro

HLE

CritPt

GPQA

AA-LCR

LiveCodeBench

BenchLM Coding score

Terminal-Bench Hard

AA Coding Index

AA-SciCode

SciCode

AA-Omniscience Hallucination Rate

AA-HLE

AA-GPQA Diamond

AA-Omniscience Accuracy

Artificial Analysis Intelligence Index

BenchLM Knowledge score

Design Arena Website

AA-IFBench

Tau2-Telecom

Claw-Eval

AA Agentic Index

Gert Labs

BenchLM Agentic score

API audit comparison

Provider examples

When to choose each model

FAQ

Related compare reports

Key Takeaways

Change comparison

DeepSeek V3.2 vs DeepSeek V4 Flash

DeepSeek V3.2

DeepSeek V4 Flash

Benchmark score comparison

Category performance

Professional benchmark details

BenchLM overall score

Output price

Blended price

Input price

MMLU-Pro

HLE

CritPt

GPQA

AA-LCR

LiveCodeBench

BenchLM Coding score

Terminal-Bench Hard

AA Coding Index

AA-SciCode

SciCode

AA-Omniscience Hallucination Rate

AA-HLE

AA-GPQA Diamond

AA-Omniscience Accuracy

Artificial Analysis Intelligence Index

BenchLM Knowledge score

Design Arena Website

AA-IFBench

Tau2-Telecom

Claw-Eval

AA Agentic Index

Gert Labs

BenchLM Agentic score

API audit comparison

Provider examples

When to choose each model