Back to Home
COMPAREPUBLIC · REFERENCE
Model comparison
Compare public specifications, then verify details on each provider's official site
Public reference data
Side-by-side comparison
Comparing 3 models · public reference only
| Parameter | DeepSeek-V4-ProView profile | Qwen3-MaxView profile | Kimi K2View profile |
|---|---|---|---|
| Core specs | |||
| Vendor | DeepSeek | Alibaba Cloud | Moonshot AI |
| Parameters | 1.6T total · 49B active (MoE) | Not publicly disclosed | Not publicly disclosed |
| MMLU (public ref.) | 88.2 (vendor-reported ref.) | 87.5 (vendor-reported ref.) | 86.9 (vendor-reported ref.) |
| Context window | 1M tokens | 128K tokens (public ref.) | 256K tokens |
| Max output | Up to 384K tokens | Per DashScope limits | Per Moonshot limits |
| Capabilities | |||
| Modalities | Text | Text, image, video, audio | Text · long-document focus |
| Tool / function calling | Supported | Supported | Supported |
| Typical strengths | Reasoning, coding, low invocation cost | Enterprise multimodal, balanced quality | Long context, document workflows |
| Access & cost | |||
| API format | OpenAI-compatible REST API | OpenAI-compatible (DashScope) | OpenAI-compatible API |
| Deployment | Cloud API | Cloud API · enterprise tiers | Cloud API |
| Languages | Chinese, English, multilingual | Chinese, English, multilingual | Chinese, English |
| Vendor-listed price | Per official pricing page | Per DashScope pricing | Per Moonshot platform |
| Knowledge cutoff | Not publicly disclosed | 2025-04 | 2025-04 |
Table summarizes public references; not an official or third-party benchmark certification.