Compare AI models

Specifications
Parameters
671B4B/30B/235B
Context window
128k256k
API pricing
Cheapest input
$0.259SiliconFlow—
Cheapest output
$0.38DeepInfra—
$71.00DeepInfra · fp4—
Benchmarks
LiveCodeBench
83.3%74.1%
Humanity's Last Exam · text only, no tools
25.1%18.2%
AIME 2025
93.1%92.3%
GPQA Diamond
82.4%81.1%
MMLU-Pro
85%84.4%
Overview
CompanyDeepSeekQwen
Release dateDec 1 2025Jul 21 2025
AccessOpen WeightOpen Weight
Model detailsView modelView model

Frequently asked questions

DeepSeek-V3.2 leads Qwen3-2507 on 5 of the 5 benchmarks they both report. Qwen3-2507 shipped 133 days before DeepSeek-V3.2, so benchmark comparisons should account for the intervening progress.

DeepSeek-V3.2 has 671B parameters, while Qwen3-2507 has 4B/30B/235B. Context windows are 128k (DeepSeek-V3.2) vs 256k (Qwen3-2507).

On LiveCodeBench, DeepSeek-V3.2 leads at 83.3% vs Qwen3-2507 at 74.1%. On Humanity's Last Exam · text only, no tools, DeepSeek-V3.2 leads at 25.1% vs Qwen3-2507 at 18.2%. On AIME 2025, DeepSeek-V3.2 leads at 93.1% vs Qwen3-2507 at 92.3%. On GPQA Diamond, DeepSeek-V3.2 leads at 82.4% vs Qwen3-2507 at 81.1%. On MMLU-Pro, DeepSeek-V3.2 leads at 85% vs Qwen3-2507 at 84.4%.