Side-by-side comparison
Pick up to 4 models. Differences are what matter for picking.
Models
| Spec | DeepSeek DeepSeek-V3.1 Verified 2026-04-15 |
|---|---|
| Description | Open MoE with hybrid thinking/non-thinking modes. Excellent price-to-performance. |
| Released | 2025-08-21 |
| Knowledge cutoff | 2024-07 |
| Context window | 128K |
| Max output | 8K |
| Input $/1M | $0.270 |
| Output $/1M | $1.10 |
| Speed tier | Standard |
| Reasoning | 4/5 |
| Coding | 5/5 |
| Multilingual | 4/5 |
| Modalities in | Text |
| Modalities out | Text |
| Tool use | Yes |
| JSON mode | Yes |
| Open weights | Yes |
| Best use cases | Coding Reasoning Agents Structured Output |
| Best personas | Startup Developer Tools Research |
| Weak at |
|
| Compliance | HIPAA-eligibleSOC 2GDPRISO 27001FedRAMPChina-based hosting raises data residency questions for EU/US enterprises. |
| Data policy | Review provider TOS; self-host weights for full control. |
| Sources |
Tip: hop over to the Recommender to describe a problem and have AI suggest the best fit from this dataset.