Claude 4.6 Sonnet vs DeepSeek-V4-Pro
Direct spec comparison of Claude 4.6 Sonnet (from Anthropic) and DeepSeek-V4-Pro (from DeepSeek). Want a 3- or 4-way comparison? Open the multi-model tool →
 Claude 4.6 SonnetAnthropic |  DeepSeek-V4-ProDeepSeek | |
|---|---|---|
| Vendor | Anthropic | DeepSeek |
| Family | Claude | DeepSeek |
| Release date | 2026-02-17 | 2026-04-22 |
| Context window | 1,000,000 tokens | 1,000,000 tokens |
| Parameters | — | 1.6T (49B active) |
| Modality | text, vision | text |
| License | proprietary | MIT |
| Source | proprietary | open weights |
| Description | Anthropic's mid-tier flagship — released 12 days after Opus 4.6 and matching most of its capabilities at ~40% the API cost ($3/$15 per 1M tokens). 1M-token context. | DeepSeek's flagship open-weight MoE. 1.6T parameters with 49B activated, 1M-token context, and a hybrid attention scheme (CSA + HCA) that delivers long-context inference at ~27% of V3.2's FLOPs. |
| Links | ||
| Benchmarks | ||
| MMLU-Pro | 85.4% | 84.2% |
| GPQA-D | 85.8% | 82.4% |
| HumanEval | 95.9% | 95.1% |
| Aider | 83.6% | 80.1% |
| AIME-25 | 87.4% | 88.6% |
| LiveCB | 73.8% | 72.4% |
More like this
Looking for a different head-to-head? Build your own comparison on the multi-model tool.
Or see all Claude 4.6 Sonnet details / DeepSeek-V4-Pro details.