DeepSeek-V3.2 is a large language model designed to harmonize high computational efficiency with strong reasoning and agentic tool-use performance. It introduces DeepSeek Sparse Attention (DSA), a fine-grained sparse attention mechanism tha…
Model details →DeepSeek: DeepSeek V3.2 vs Inception: Mercury 2
DeepSeek: DeepSeek V3.2 wins on 2 of 6 axes — pricing and capability skew in its favour for most workloads.
Mercury 2 is an extremely fast reasoning LLM, and the first reasoning diffusion LLM (dLLM). Instead of generating tokens sequentially, Mercury 2 produces and refines multiple tokens in parallel, achieving >1,000 tokens/sec on standard GPUs.…
Model details →Side-by-side comparison
| Capability | DeepSeek: DeepSeek V3.2 | Inception: Mercury 2 | Winner |
|---|---|---|---|
| Context window Maximum number of input tokens the model can attend to in a single request. | 164K | 128K | 🏆 DeepSeek: DeepSeek V3.2 |
| Input price (per 1M) Cost per million input tokens billed by the provider. | $0.26 | $0.25 | 🏆 Inception: Mercury 2 |
| Output price (per 1M) Cost per million output tokens billed by the provider. | $0.38 | $0.75 | 🏆 DeepSeek: DeepSeek V3.2 |
| Tool / function calling First-class support for emitting structured tool calls. | Yes | Yes | Tie |
| Vision input Accepts image inputs alongside text. | — | — | Tie |
| Reasoning mode Internal chain-of-thought / extended-thinking support. | Yes | Yes | Tie |
Frequently asked questions
Is DeepSeek: DeepSeek V3.2 better than Inception: Mercury 2?
DeepSeek: DeepSeek V3.2 wins on 2 of 6 axes — pricing and capability skew in its favour for most workloads.
What's the price difference between DeepSeek: DeepSeek V3.2 and Inception: Mercury 2?
Input: $0.26 vs $0.25 per 1M tokens. Output: $0.38 vs $0.75 per 1M tokens.
What context windows do DeepSeek: DeepSeek V3.2 and Inception: Mercury 2 support?
DeepSeek: DeepSeek V3.2 supports up to 164K tokens. Inception: Mercury 2 supports up to 128K tokens.
Do both DeepSeek: DeepSeek V3.2 and Inception: Mercury 2 support tool calling?
DeepSeek: DeepSeek V3.2: yes. Inception: Mercury 2: yes.