DeepSeek V3.2
DeepSeek · 2 thinking configurations
Intelligence21.5
Input / 1M$0.28
Output / 1M$0.42
Context128K
Tariff: DeepSeek · Source: artificial-analysis · USD per million tokens
Thinking configuration
What does more thinking buy you?
Compare this model’s measured thinking configurations.
Scroll the chart ↔ · tap a point for details
Efficient frontier: best measured score at each cost2 measured configurations · 0 without both values
Selected: thinking thinking · Click a point to switch configuration
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...
Intelligence IndexArtificial Analysis composite index, not a percentage. Different versions of the index are not directly comparable.
21.5 pointsGPQA DiamondGraduate-level scientific questions. Measures specialist reasoning rather than web-search quality.
84.0 %Humanity’s Last ExamDifficult questions across many disciplines, compared under the same evaluation setup.
24.6 %Terminal-Bench 2.1Agent tasks in a terminal environment. Results also depend on the evaluation harness.
46.8 %τ²-BenchTool use and interaction in assistance scenarios. Does not replace an evaluation of your integration.
90.6 %IFBenchVerifiable instruction following. A proxy for precision rather than a benchmark of creative style.
60.7 %Omniscience accuracyFactual accuracy in Artificial Analysis's Omniscience evaluation.
33.0 %Specifications from the catalog
Inputtext
Outputtext
Tool callingYes
Structured outputYes
SpeedNot available
First tokenNot available
Maximum output65.5K
Catalog listing date2025-12-01