GPT OSS 20B vs Qwen: Qwen3 235B A22B Thinking 2507
Compare GPT OSS 20B by OpenAI against Qwen: Qwen3 235B A22B Thinking 2507 by Qwen, in 27 community votes, qwen: qwen3 235b a22b thinking 2507 wins 68% of head-to-head duels, context windows of 131K vs 131K, tested across 54 shared challenges.
Web Design: Qwen: Qwen3 235B A22B Thinking 2507 wins 57% of votes
Reasoning: Qwen: Qwen3 235B A22B Thinking 2507 wins 80% of votes
Conversation: Qwen: Qwen3 235B A22B Thinking 2507 wins 100% of votes
Image Generation: Qwen: Qwen3 235B A22B Thinking 2507 wins 75% of votes
Analysis: GPT OSS 20B wins 100% of votes
Our Verdict
Qwen: Qwen3 235B A22B Thinking 2507Winner
GPT OSS 20BRunner-up
Pick Qwen: Qwen3 235B A22B Thinking 2507. In 27 blind votes, Qwen: Qwen3 235B A22B Thinking 2507 wins 68% of the time. That's not luck.
Pick GPT OSS 20B for Analysis. Pick Qwen: Qwen3 235B A22B Thinking 2507 for Conversation, Reasoning, Image Generation. GPT OSS 20B is 6.0x cheaper per token — worth considering if cost matters.
Clear winner
API pricing
Cost per 1M tokens
GPT OSS 20B
Input
$0.02
5.5× cheaper
Output
$0.10
6.0× cheaper
Qwen: Qwen3 235B A22B Thinking 2507
Input
$0.11
Output
$0.60
GPT OSS 20B is cheaper on both — 5.5× input, 6.0× output
Writing DNA
Style Comparison
Similarity
39%
Qwen: Qwen3 235B A22B Thinking 2507 uses 5.3x more emoji
GPT OSS 20B
Qwen: Qwen3 235B A22B Thinking 2507
54%Vocabulary56%
17wSentence Length14w
0.26Hedging0.37
5.9Bold5.7
3.3Lists4.0
0.10Emoji0.54
0.75Headings0.70
0.32Transitions0.14
Based on 21 + 22 text responses
Research
What we learned reading every model
FAQ
