The AI Cold War's Newest Weapon: Stealing the Teacher's Answer Key · Judge: anthropic:claude-opus-4-8 · 2026-05-30T15:43:59
· ✅ saved (e5c57c) · 14 models
🔬 View outputs side-by-side → all verdicts →
| # | Model | Facts | Analogy | Age-fit | Clarity | Quiz | Glossary | Passage | Overall | Cost | Value | Notes |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 🥇 | openrouter:deepseek/deepseek-v4-pro
view ↗ |
10 | 9 | 9 | 9 | 9 | 9 | 9 | 9 | – | – | Excellent fact fidelity with the Tu-4/B-29 Cold War analogy sharing genuine mechanism (reverse-engineering that doesn't fully match original). Ten well-constructed quiz questions with proper evidence-pairing. Rich glossary and a polished, original rewritten passage. |
| 🥈 | openrouter:minimax/minimax-m2.7
view ↗ |
9 | 8 | 9 | 9 | 9 | 9 | 9 | 9 | – | – | Accurate and well-organized with strong 'splinternet' framing. Quiz questions tie tightly to the rewritten passage and evidence-pairing works. 'Surreptitious' question references a word not in the rewritten passage but the quotes are real. |
| 🥉 | openrouter:moonshotai/kimi-k2.6
view ↗ |
9 | 9 | 9 | 9 | 9 | 9 | 9 | 9 | – | – | Outstanding music-conservatory and pharma-abstract analogies that precisely match the mechanism. Faithful, complete rewritten passage incorporating McGuire and embassy details. Strong evidence-pairing quiz with clear trap explanations. |
| 4 | openrouter:mistralai/mistral-large-2512
view ↗ |
9 | 8 | 9 | 9 | 8 | 8 | 9 | 8 | – | – | Strong, faithful coverage with vivid analogies (answer key, marathon sweatband). The 'surreptitious' vocab question asks about a word not in the rewritten passage, a minor flaw. Otherwise clear and well-pitched. |
| 5 | openrouter:google/gemini-3.1-flash-lite
view ↗ |
9 | 8 | 8 | 9 | 8 | 8 | 9 | 8 | – | – | Clean, faithful, with apt chef and capture-the-flag analogies. Good ten-question quiz with evidence pairing. Solid rewritten passage. Slightly less distinctive than the top finishers. |
| 6 | openrouter:z-ai/glm-5-turbo
view ↗ |
9 | 8 | 9 | 9 | 7 | 9 | 7 | 8 | – | – | Vivid music-producer and scouting-report analogies, strong glossary. However the 'sharp' vocab question (Q4) tests a phrase the model fabricated as if from the source, and the rewritten passage is truncated. Otherwise excellent. |
| 7 | openrouter:x-ai/grok-4.3
view ↗ |
9 | 7 | 8 | 8 | 8 | 8 | 8 | 8 | – | – | Accurate and complete with reasonable reverse-engineering and scouting analogies. Good quiz with evidence pairing keyed to the rewritten passage. Slightly more compact and less inventive than top entries. |
| 8 | openrouter:openai/gpt-5.4-nano
view ↗ |
9 | 8 | 8 | 8 | 8 | 9 | 8 | 8 | – | – | Careful, honest framing (consistently flags 'the article says'), strong export-controls/lock and scouting analogies. Good glossary and complete passage. Slightly hedged tone but very faithful to source. |
| 9 | openrouter:stepfun/step-3.7-flash
view ↗ |
8 | 8 | 8 | 7 | 8 | 8 | 7 | 7 | – | – | Solid analogies and good SAT-style traps, but the rewritten passage is truncated mid-sentence, hurting completeness. Adds 'state-linked' framing not strictly in source. 'Surreptitious' question references content not in the truncated passage. |
| 10 | openrouter:qwen/qwen3-max-thinking
view ↗ |
9 | 8 | 8 | 8 | 5 | 8 | 9 | 7 | – | – | Strong rewritten passage and good analogies, but only ONE quiz question is provided despite the format expecting a full set, a major shortfall. Otherwise accurate and well-pitched. |
| 11 | openrouter:meta-llama/llama-4-maverick
view ↗ |
8 | 7 | 8 | 8 | 8 | 7 | 7 | 7 | – | – | Generic but functional analogies (arms race, doping). Quiz is solid. The rewritten passage adds 'urgent respon[se]' and is truncated; states the memo focuses on three companies somewhat overconfidently. Reasonable but not standout. |
| 12 | openrouter:bytedance-seed/seed-2.0-lite
view ↗ |
8 | 7 | 8 | 8 | 7 | 8 | 8 | 7 | – | – | Good analogies (MIT exam keys, Twitch scouting). Q4 tests 'principal'—a word that appears in source as 'principally,' a slight stretch. Otherwise accurate, complete passage and decent quiz with evidence pairing. |
| 13 | openrouter:amazon/nova-pro-v1
view ↗ |
7 | 5 | 6 | 6 | 6 | 6 | 6 | 6 | – | – | Competent but generic. The chess and doping analogies are weak/mechanism-mismatched. Quiz questions sometimes cite phrases ('global power dynamics') that exist in the explainer rather than the source. Thinner glossary and shorter passage. |
| 14 | openrouter:baidu/ernie-4.5-vl-424b-a47b
view ↗ |
7 | 5 | 7 | 6 | 5 | 6 | 6 | 5 | – | – | Only three quiz questions and a thin three-term glossary. Q2 marks 'replicate' over the source's literal definition awkwardly. Analogies are generic. Shorter and less rigorous than peers. |