DeepSeek V3 vs Qwen3 72B Instruct

Capability, benchmark, and verified hosting price comparison.

Quick verdict

Benchmark wins: DeepSeek V3— leads on 2 of 2 shared benchmarks
Cheapest hosting: DeepSeek V3— $0.477 vs $0.900 / 1M blended tokens
Context window: Qwen3 72B Instruct— 131,072 tokens vs 65,536

Derived from public benchmarks and verified hosting prices. Benchmark wins count only where both models have a score; direction respects each benchmark’s higher-is-better flag.

DeepSeek V3

by DeepSeek AI · DeepSeek

Context: 65,536 tokens
Open weight: Yes
License: DeepSeek License v1.0
Params: 671.0B
Modalities: text

Qwen3 72B Instruct

by Alibaba · Qwen

Context: 131,072 tokens
Open weight: Yes
License: Apache 2.0
Params: 72.0B
Modalities: text

Benchmarks

Benchmark	DeepSeek V3	Qwen3 72B Instruct
HumanEval	0.825	0.821
MMLU	0.887	0.859

Cheapest verified hosting

DeepSeek V3 cheapest

DeepSeek

$0.477/ 1M tokens (blended)

Input: $0.270
Output: $1.100
Context: 65,536 tok
Latency p50: 480 ms

Qwen3 72B Instruct cheapest

Together.ai

$0.900/ 1M tokens (blended)

Input: $0.900
Output: $0.900
Context: 131,072 tok
Latency p50: 220 ms