Home › Free Underdog Scorecard
Free Underdog Scorecard (1-Page PDF)
What you get
One page: DeepSeek-R1 vs OpenAI o1 side by side (price per 1M output tokens as of January 2025, and two benchmark scores as reported by DeepSeek), the sources, and a blank scorecard to compare any two AI tools yourself. Free, no email, no sign-up.
Download the Underdog Scorecard (PDF)
What's on it
- Output price per 1M tokens, January 2025: DeepSeek-R1 $2.19, OpenAI o1 $60.
- AIME 2024 (math): R1 79.8, o1 79.2. GPQA Diamond (science): o1 75.7, R1 71.5. Scores are self-reported by DeepSeek and not independently verified.
- Model versions, price dates and source links.
- A blank template (Model / Price / Performance / Speed / Best at / Sources) and a "Which one would you choose?" box.
New: Episode 2 (page 2)
- Qwen3-8B (trained with SFT, then RL) vs Llama-4-Maverick (402B total, 17B active; prompted). ToolBench G3 tool selection, exact match, 2,124 instructions: 41.5% vs 30.6% (Maverick's best of four prompting setups).
- On a different benchmark, TaskBench, Maverick scored higher: 80.6% vs 72.8%.
- Maverick was also the teacher model used to create the small model's training examples.
- Source: Zhou et al. (Meta; Carnegie Mellon University), "Training Small Language Models for Transferable Tool Selection", arXiv preprint, Oct 22, 2026 (not peer-reviewed).
Honest notes
Prices and models change. Check today's official pages before you decide. No overall winner is declared; that's your call.
Names are trademarks of their owners. Peace Call is not affiliated with or endorsed by OpenAI or DeepSeek. Made with AI: Mr. Peace is an AI character with an AI voice.