Compare the major reasoning models: o1, o3, DeepSeek-R1, Claude 3.7, Gemini 2.0 Flash Thinking.
LLM & AILevel: IntermediateKey Models & Comparisons
Compare the major reasoning models: o1, o3, DeepSeek-R1, Claude 3.7, Gemini 2.0 Flash Thinking.
Hints
- Identify the stage: pretraining, adaptation, generation or evaluation.
- Name the inputs, constraints and a case where the approach fails.
Answer
| Model | Provider | Open Source | Thinking Visible | Key Strength |
|---|---|---|---|---|
| o1 | OpenAI | No | No (hidden) | First major reasoning model; strong math/code |
| o3 | OpenAI | No | No | State-of-art on benchmarks; high cost |
| o4-mini | OpenAI | No | No | Efficient reasoning; good cost/perf ratio |
| DeepSeek-R1 | DeepSeek | Yes | Yes | Open weights; matches o1 at fraction of cost |
| Claude 3.7 Sonnet | Anthropic | No | Yes (optional) | Extended thinking; strong coding |
| Gemini 2.0 Flash Thinking | No | Yes | Fast; multimodal reasoning | |
| Qwen QwQ | Alibaba | Yes | Yes | Strong math; open weights |
DeepSeek-R1's significance: it proved RLVR reasoning training works with open-source models, democratizing the technique. R1's distilled variants (1.5B–70B) brought reasoning capabilities to consumer hardware.
Source and licence: JustInCache & contributors — Awesome AI Interviews · Apache-2.0