deepseek/deepseek-v3.2
DeepSeek V3.2
DeepSeek‑V3.2 from DeepSeek harmonizes high computational efficiency with superior reasoning and agent performance. It builds on three main techniques: DeepSeek Sparse Attention for long‑context efficiency, a scalable reinforcement learning framework, and a large‑scale agentic task synthesis pipeline. This model excels at long-context reasoning and agentic tasks, efficiently handling extended inputs while maintaining strong accuracy. Its sparse attention design enables it to process complex, multi-step workflows without excessive compute costs. Overall, DeepSeek‑V3.2 targets long‑context reasoning, tool‑using agents, and efficient deployment in production environments.
Price comparison
USD per one million tokens, using current provider data.
Provider latency
Last-hour p50 in milliseconds; lower is better.
Provider uptime
Reported availability over the last 24 hours; not a contractual SLA.
Inference providers
Compare current pricing, performance, and privacy across every available provider.
| Provider | Maximum context | Max output | p50 latency | p50 throughput | 24h uptime | Input / 1M tokens | Output / 1M tokens | Privacy |
|---|---|---|---|---|---|---|---|---|
Bedrock | 128K | 8K | 1,025 ms | 28.0 t/s | 99.998% | $0.62 | $1.85 | |
Deepinfra | 163.8K | 8K | 659 ms | 16.0 t/s | 100.000% | $0.26 | $0.38 | |
Novita | 163.8K | 65.5K | 1,540 ms | 27.0 t/s | 100.000% | $0.28 | $0.42 |
What can this model do?
Available modalities and capabilities across inference providers.
Technical profile
- Model type
- Language model
- Release date
- Dec 1, 2025
- Knowledge cutoff
- 2024-07
- Regions
- —
- API specifications
- V2, V3, V4
- Temperature control
- Supported
Model identifier
deepseek/deepseek-v3.2Related models
More options from the same creator or model category.
