PyyanThe AI Index

Pyyan / Compare / Qwen3-235B-A22B vs DeepSeek V4-Pro vs Llama 5

Qwen3-235B-A22B vs DeepSeek V4-Pro vs Llama 5

3 of 5

Open-Weight Models · verified 13 Aug 2026

×Qwen3-235B-A22BAlibabacurrent
×DeepSeek V4-ProDeepSeekcurrent
×Llama 5Metacurrent
2 slots left
SpecificationQwen3-235B-A22BDeepSeek V4-ProLlama 5
SummaryCheapest capable model, most permissive licence.Best all-round open model of 2026.Open-weight flagship with recursive self-improvement.
LicenceApache-2.0MITLlama licence
Parameters235B MoE—600B+
CategoryOpen-Weight ModelsOpen-Weight ModelsOpen-Weight Models
OfficialAlibaba ↗DeepSeek ↗Meta ↗

Highlighted rows are where these differ.

Qwen3-235B-A22B

  • No licence conditions whatsoever
  • 22B active of 235B total

Best for commercial self-hosting.

Full spec sheet →

DeepSeek V4-Pro

  • Tops open leaderboards on agentic coding and reasoning
  • MIT licence — no conditions
  • Pioneered aggressive cache-hit pricing at ~$0.07/M

Best for everything open.

Full spec sheet →

Llama 5

  • Refines its own reasoning and generates synthetic training data

Best for open frontier.

Full spec sheet →