Model comparison
DeepSeek R1 vs Claude Sonnet 4.6
DeepSeek
DeepSeek R1
The open-weights reasoning model that reset the cost curve.
Anthropic
Claude Sonnet 4.6
The pragmatic default - Claude quality without Opus pricing.
Specs
| Metric | DeepSeek R1 | Claude Sonnet 4.6 |
|---|---|---|
| Context window | 164K tokens | 1M tokens↑ |
| Input $/1M tokens | $0.700↑ | $3.00 |
| Output $/1M tokens | $2.50↑ | $15.00 |
| Modalities | Text | Text · Image · File |
| Open weights | Yes | No |
| Released | Jan 2025 | - |
Capability differences
| Capability | DeepSeek R1 | Claude Sonnet 4.6 |
|---|---|---|
| Tool use | No | Yes |
| Vision | No | Yes |
| Prompt caching | No | Yes |
| Open weights | Yes | No |
Key differences
- -DeepSeek R1 offers 164K context; Claude Sonnet 4.6 offers 1M.
- -Input pricing: DeepSeek R1 at $0.700/1M vs Claude Sonnet 4.6 at $3.00/1M.
- -DeepSeek R1 is open weights; Claude Sonnet 4.6 is proprietary.