Compare · ModelsLive · 2 picked · head to head

DeepSeek-V2 (MoE-236B, May 2024) vs Qwen2.5 72B Instruct

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

Winner summary

DeepSeek-V2 (MoE-236B, May 2024) wins 4 of 7 shared benchmarks. Leads in knowledge.

Category leads
knowledge·DeepSeek-V2 (MoE-236B, May 2024)reasoning·Qwen2.5 72B Instruct

Head to head · from the data · data as of 2026-11-22

  • Benchmarks · DeepSeek-V2 (MoE-236B, May 2024) leads on 4 of 7 shared benchmarks (HellaSwag, PIQA, TriviaQA, Winogrande), Qwen2.5 72B Instruct on 3 (ARC AI2, BBH, MMLU)
  • Cheapest provider · Qwen2.5 72B Instruct at DeepInfra $0.29/M input (of 2 offers) as of 2026-11-22
  • Open weights · DeepSeek-V2 (MoE-236B, May 2024) and Qwen2.5 72B Instruct both have open weights
  • Price history · Qwen2.5 72B Instruct OpenRouter input list price had no move of 5% or more in the 90 days to 2026-11-22

Sources · scores from public leaderboards, list prices and provider offers from OpenRouter, price moves from the BenchGecko changelog (5% or more, held in the next daily check), Gecko tests run by BenchGecko.

Hype vs Reality
DeepSeek-V2 (MoE-236B, May 2024)
#8 by perf·no signal
QUIET
Qwen2.5 72B Instruct
#144 by perf·#1 by attention
DESERVED
Best value
DeepSeek-V2 (MoE-236B, May 2024)
n/a
no price
Qwen2.5 72B Instruct
129.7 pts/$
$0.38/M
Vendor risk
One or more vendors flagged
DeepSeek logo
DeepSeek
$3.4B·Tier 1
Higher risk
Alibaba Qwen logo
Alibaba (Qwen)
$293.0B·Tier 1
Low risk
Head to head
DeepSeek-V2 (MoE-236B, May 2024)Qwen2.5 72B Instruct
ARC AI2
Qwen2.5 72B Instruct leads by +3.1
AI2 Reasoning Challenge · tests grade-school level science knowledge with multiple-choice questions requiring reasoning beyond simple retrieval.
DeepSeek-V2 (MoE-236B, May 2024)
89.6
Qwen2.5 72B Instruct
92.7
BBH
Qwen2.5 72B Instruct leads by +1.3
BIG-Bench Hard · a curated subset of 23 challenging tasks from BIG-Bench where language models previously failed to outperform average humans.
DeepSeek-V2 (MoE-236B, May 2024)
71.7
Qwen2.5 72B Instruct
73.1
HellaSwag
DeepSeek-V2 (MoE-236B, May 2024) leads by +3.1
HellaSwag · tests commonsense reasoning by asking models to predict the most plausible continuation of everyday scenarios.
DeepSeek-V2 (MoE-236B, May 2024)
82.8
Qwen2.5 72B Instruct
79.7
MMLU
Qwen2.5 72B Instruct leads by +9.2
Massive Multitask Language Understanding · 57 subjects spanning STEM, humanities, social sciences, and more. The standard benchmark for broad knowledge.
DeepSeek-V2 (MoE-236B, May 2024)
71.2
Qwen2.5 72B Instruct
80.4
PIQA
DeepSeek-V2 (MoE-236B, May 2024) leads by +2.6
PIQA (Physical Interaction QA) · tests intuitive physical reasoning by asking models to select the correct approach for everyday physical tasks.
DeepSeek-V2 (MoE-236B, May 2024)
67.8
Qwen2.5 72B Instruct
65.2
TriviaQA
DeepSeek-V2 (MoE-236B, May 2024) leads by +8.1
TriviaQA · reading comprehension benchmark with trivia questions, requiring models to find and reason over evidence from provided documents.
DeepSeek-V2 (MoE-236B, May 2024)
80.0
Qwen2.5 72B Instruct
71.9
Winogrande
DeepSeek-V2 (MoE-236B, May 2024) leads by +8.0
WinoGrande · large-scale commonsense reasoning benchmark where models must resolve ambiguous pronouns in carefully constructed sentence pairs.
DeepSeek-V2 (MoE-236B, May 2024)
72.6
Qwen2.5 72B Instruct
64.6
Full benchmark table
BenchmarkDeepSeek-V2 (MoE-236B, May 2024)Qwen2.5 72B Instruct
ARC AI2
AI2 Reasoning Challenge · tests grade-school level science knowledge with multiple-choice questions requiring reasoning beyond simple retrieval.
89.692.7
BBH
BIG-Bench Hard · a curated subset of 23 challenging tasks from BIG-Bench where language models previously failed to outperform average humans.
71.773.1
HellaSwag
HellaSwag · tests commonsense reasoning by asking models to predict the most plausible continuation of everyday scenarios.
82.879.7
MMLU
Massive Multitask Language Understanding · 57 subjects spanning STEM, humanities, social sciences, and more. The standard benchmark for broad knowledge.
71.280.4
PIQA
PIQA (Physical Interaction QA) · tests intuitive physical reasoning by asking models to select the correct approach for everyday physical tasks.
67.865.2
TriviaQA
TriviaQA · reading comprehension benchmark with trivia questions, requiring models to find and reason over evidence from provided documents.
80.071.9
Winogrande
WinoGrande · large-scale commonsense reasoning benchmark where models must resolve ambiguous pronouns in carefully constructed sentence pairs.
72.664.6
Pricing · per 1M tokens · projected $/mo at 10M tokens
ModelInputOutputContextProjected $/mo
DeepSeek logoDeepSeek-V2 (MoE-236B, May 2024)n/an/an/an/a
Alibaba Qwen logoQwen2.5 72B Instruct$0.36$0.4033K tokens (~16 books)$3.70
DeepSeek-V2 (MoE-236B, May 2024) leads on 4 of 7 shared benchmarks (HellaSwag, PIQA, TriviaQA, Winogrande), Qwen2.5 72B Instruct on 3 (ARC AI2, BBH, MMLU)