DeepSeek V3.2 Exp pricing & benchmarks
Released Sep 29, 2025 · served by 3 providers
What DeepSeek V3.2 Exp is
DeepSeek-V3.2-Exp is an experimental large language model released by DeepSeek as an intermediate step between V3.1 and future architectures. It introduces DeepSeek Sparse Attention (DSA), a fine-grained sparse attention mechanism...
- Model ID:
deepseek/deepseek-v3.2-exp - Modality: text->text
- Knowledge cutoff: 2025-07-31
- Tool calling: supported · Structured output: supported
- Weights:
deepseek-ai/DeepSeek-V3.2-Expon Hugging Face
Providers and prices for DeepSeek V3.2 Exp
The same model costs different amounts depending on who serves it. Prices are per 1M tokens, cheapest first. Uptime and throughput are OpenRouter's rolling measurements, not vendor claims.
| Provider | Input | Output | Context | Quant | Uptime 24h | Throughput |
|---|---|---|---|---|---|---|
| Novitacheapest | $0.270 | $0.410 | 164K | fp8 | 98.8% | — |
| SiliconFlow | $0.270 | $0.410 | 164K | fp8 | 89.7% | — |
| AtlasCloud | $0.270 | $0.410 | 164K | fp8 | 98.8% | — |
DeepSeek V3.2 Exp benchmark results
Design Arena head-to-head results, as reported through the OpenRouter model API.
| Category | Arena | Rank | Elo | Win rate |
|---|---|---|---|---|
| 3d | models | #44 | 1206 | 56.4% |
| uicomponent | models | #46 | 1203 | 53.3% |
| gamedev | models | #48 | 1201 | 53.1% |
| codecategories | models | #51 | 1201 | 54.2% |
| website | models | #53 | 1204 | 54.2% |
| dataviz | models | #58 | 1181 | 50.6% |
| svg | models | #58 | 1082 | 42% |
FAQ
How much does the DeepSeek V3.2 Exp API cost?
$0.270 per 1M input tokens and $0.410 per 1M output tokens. A typical workload of 1M input + 300K output tokens costs about $0.39.
Is DeepSeek V3.2 Exp free?
No. It is a paid model starting at $0.270 per 1M input tokens, though some providers offer trial credits.
Which provider is cheapest for DeepSeek V3.2 Exp?
Novita at $0.270 input / $0.410 output per 1M tokens (fp8 quantization).
Can I self-host DeepSeek V3.2 Exp?
Yes — weights are published as deepseek-ai/DeepSeek-V3.2-Exp on Hugging Face, so you can run it on your own hardware.