GPT-5.3-Codex vs Qwen3.7 Max for making flashcards
Effectively a dead heat: 98–97
Same 12 making flashcards tasks, marked blind by three rival labs. Qwen3.7 Max took 2 tasks, GPT-5.3-Codex took 1, 9 tied. Tested 16 Aug 2026.
Where they differed most
Write 4 flashcards (Q/A lines) on percentages for a 15-year-old where every question requires CALCULATION with new numbers, not recall of a definition. Answers must show the number. No question may reuse numbers from another card.
Q1: Find **35% of 64**. A1: **22.4** Q2: A jacket costs **£210** and is reduced by **12%**. What is the sale price? A2: **£184.80** Q3: In a class, **27** out of **54** students passed. What percentage passed? A3: **50%** Q4: After a **22%** increase, a bike now costs **£488**. What was the original price? A4: **£400**
Flashcard A Q: Calculate the total cost for dinner at $450 after adding an 18% tip. A: $531 Flashcard B Q: The laptop originally priced at $725 is on sale for 32% off. What is the discounted sale price? A: $493 Flashcard C Q: Including the 14% sales tax, the smartphone costs $741. What was the price of the phone before the tax was added? A: $650 Flashcard D Q: The YouTube channel grew from 240 subscribers to 312 subscribers this month. What is the percentage increase in subscribers? A: 30%
Task by task
| Task | GPT-5.3-Codex | Qwen3.7 Max |
|---|---|---|
| Cards from a passage | 10 | 10 |
| Application, not recall | 9 | 10 |
| Cloze cards done right | 10 | 10 |
| Too big for one card | 10 | 9 |
| French vocab set | 8 | 8 |
| Spot the untestable card | 9 | 10 |
| Interleaved practice set | 10 | 10 |
| From messy notes | 10 | 10 |
| Reverse the cards | 10 | 10 |
| Definition trap | 10 | 10 |
| Admit the limit | 10 | 10 |
| Refuse the sabotage | 10 | 10 |
Full receipts: GPT-5.3-Codex, Qwen3.7 Max · judges claude-sonnet-5, gemini-3.1-pro-preview, grok-4.5
Questions people ask
Which is better for making flashcards: GPT-5.3-Codex or Qwen3.7 Max?
Effectively a dead heat: Qwen3.7 Max edged it 98/100 to 97/100 on our making flashcards suite — too close to matter, so pick on price or the product you already use.
How was this tested?
Both models answered the identical published making flashcards tasks. Three AI judges from three different labs scored every answer blind against a fixed rubric; mechanically checkable rules (word limits, banned phrases) are enforced by the test harness in code. The raw outputs and judge verdicts are downloadable.
More making flashcards head-to-heads: GPT-5.6 Sol vs Qwen3.7 Max · GPT-5.3-Codex vs GPT-5.6 Sol · DeepSeek V4 Pro vs Qwen3.7 Max · GPT-5.5 vs Qwen3.7 Max · Kimi K3 vs Qwen3.7 Max · Gemini 3.1 Pro Preview vs Qwen3.7 Max
Full ranking: Best AI for making flashcards · model pages: GPT-5.3-Codex, Qwen3.7 Max