Compare · ModelsLive · 2 picked · head to head

DeepSeek V3.2 vs DeepSeek V4 Flash

Side by side · benchmarks, pricing, and signals you can act on.

Winner summary

DeepSeek V4 Flash wins 9 of 12 shared benchmarks. Leads in arena · coding · reasoning.

Category leads
speed·DeepSeek V3.2arena·DeepSeek V4 Flashcoding·DeepSeek V4 Flashreasoning·DeepSeek V4 Flashlanguage·DeepSeek V4 Flashmath·DeepSeek V4 Flashknowledge·DeepSeek V4 Flash
Hype vs Reality
DeepSeek V3.2
#108 by perf·no signal
QUIET
DeepSeek V4 Flash
#31 by perf·no signal
QUIET
Best value
2.8x better value than DeepSeek V3.2
DeepSeek V3.2
191.0 pts/$
$0.27/M
DeepSeek V4 Flash
533.3 pts/$
$0.13/M
Vendor risk
One or more vendors flagged
DeepSeek logo
DeepSeek
$3.4B·Tier 1
Higher risk
DeepSeek logo
DeepSeek
$3.4B·Tier 1
Higher risk
Head to head
DeepSeek V3.2DeepSeek V4 Flash
Artificial Analysis · Agentic Index
DeepSeek V3.2 leads by +21.9
Artificial Analysis Agentic Index · a composite score measuring how well a model performs in agentic workflows · multi-step tool use, planning, error recovery, and autonomous task completion. Aggregates results from multiple agentic benchmarks including SWE-bench, tool-use tests, and planning evaluations. The canonical single-number metric for "how good is this model as an agent?"
DeepSeek V3.2
52.9
DeepSeek V4 Flash
31.1
Artificial Analysis · Coding Index
DeepSeek V4 Flash leads by +19.5
Artificial Analysis Coding Index · a composite score that aggregates performance across multiple coding benchmarks into a single index. Tracks code generation quality, debugging ability, multi-language competence, and real-world software engineering tasks. Used by Artificial Analysis to rank model coding capability in a normalized, comparable format. Useful for developers choosing between models for coding-heavy workloads.
DeepSeek V3.2
36.7
DeepSeek V4 Flash
56.2
Artificial Analysis · Quality Index
DeepSeek V3.2 leads by +1.4
DeepSeek V3.2
41.7
DeepSeek V4 Flash
40.3
Chatbot Arena Elo · Overall
DeepSeek V4 Flash leads by +8.9
DeepSeek V3.2
1425.0
DeepSeek V4 Flash
1433.8
LiveBench · Agentic Coding
DeepSeek V4 Flash leads by +3.3
DeepSeek V3.2
46.7
DeepSeek V4 Flash
50.0
LiveBench · Coding
DeepSeek V3.2 leads by +6.5
DeepSeek V3.2
75.7
DeepSeek V4 Flash
69.2
LiveBench · Data Analysis
DeepSeek V4 Flash leads by +23.0
DeepSeek V3.2
45.0
DeepSeek V4 Flash
68.0
LiveBench · If
DeepSeek V4 Flash leads by +40.1
DeepSeek V3.2
23.1
DeepSeek V4 Flash
63.1
LiveBench · Language
DeepSeek V4 Flash leads by +5.9
DeepSeek V3.2
64.2
DeepSeek V4 Flash
70.1
LiveBench · Mathematics
DeepSeek V4 Flash leads by +15.7
DeepSeek V3.2
64.0
DeepSeek V4 Flash
79.7
LiveBench · Overall
DeepSeek V4 Flash leads by +15.4
DeepSeek V3.2
51.8
DeepSeek V4 Flash
67.3
LiveBench · Reasoning
DeepSeek V4 Flash leads by +26.3
DeepSeek V3.2
44.3
DeepSeek V4 Flash
70.6
Full benchmark table
BenchmarkDeepSeek V3.2DeepSeek V4 Flash
Artificial Analysis · Agentic Index
Artificial Analysis Agentic Index · a composite score measuring how well a model performs in agentic workflows · multi-step tool use, planning, error recovery, and autonomous task completion. Aggregates results from multiple agentic benchmarks including SWE-bench, tool-use tests, and planning evaluations. The canonical single-number metric for "how good is this model as an agent?"
52.931.1
Artificial Analysis · Coding Index
Artificial Analysis Coding Index · a composite score that aggregates performance across multiple coding benchmarks into a single index. Tracks code generation quality, debugging ability, multi-language competence, and real-world software engineering tasks. Used by Artificial Analysis to rank model coding capability in a normalized, comparable format. Useful for developers choosing between models for coding-heavy workloads.
36.756.2
Artificial Analysis · Quality Index
41.740.3
Chatbot Arena Elo · Overall
1425.01433.8
LiveBench · Agentic Coding
46.750.0
LiveBench · Coding
75.769.2
LiveBench · Data Analysis
45.068.0
LiveBench · If
23.163.1
LiveBench · Language
64.270.1
LiveBench · Mathematics
64.079.7
LiveBench · Overall
51.867.3
LiveBench · Reasoning
44.370.6
Pricing · per 1M tokens · projected $/mo at 10M tokens
ModelInputOutputContextProjected $/mo
DeepSeek logoDeepSeek V3.2$0.21$0.32131K tokens (~66 books)$2.41
DeepSeek logoDeepSeek V4 Flash$0.08$0.171.0M tokens (~524 books)$1.05