DeepSeek: DeepSeek V3.2
by DeepSeekDeepSeek: DeepSeek V3.2 is a large language model from DeepSeek. It costs $0.269 per million input tokens and $0.400 per million output tokens. Its context window is 164K tokens.
- Input / 1M tokens
- $0.269
- Output / 1M tokens
- $0.400
- Cached input / 1M
- $0.134
- Context window
- 164K
On repeated prefixes
tokens
Who serves it cheapest
15 hosts serve DeepSeek V3.2. Same weights, same API - the price difference is pure margin and routing.
| Provider | Input / 1M | Output / 1M | Context | Throughput | Uptime 24h |
|---|---|---|---|---|---|
| GMICloudCheapestfp8 | $0.209 | $0.310 | 164K | - | 99.9% |
| StreamLakefp8 | $0.214 | $0.322 | 128K | - | 98.7% |
| DeepInfrafp4 | $0.260 | $0.380 | 164K | - | 99.2% |
| AtlasCloudfp8 | $0.260 | $0.380 | 164K | - | 99.6% |
| Venice | $0.268 | $0.390 | 160K | - | 99.1% |
| SiliconFlowfp8 | $0.259 | $0.420 | 164K | - | 97.2% |
| Novitafp8 | $0.269 | $0.400 | 164K | - | 99.9% |
| Baidufp8 | $0.280 | $0.420 | 131K | - | 100.0% |
| DigitalOcean | $0.300 | $0.960 | 164K | - | 93.9% |
| Alibabafp8 | $0.370 | $1.11 | 131K | - | 99.1% |
| Friendli | $0.500 | $1.50 | 164K | - | 100.0% |
| $0.560 | $1.68 | 164K | - | 99.2% | |
| Phala | $1.00 | $1.00 | 164K | - | 98.7% |
| Mara | $3.00 | $4.50 | 33K | - | 79.5% |
| SambaNova | $3.00 | $4.50 | 33K | - | 90.8% |
The spread between GMICloud and SambaNova is 14× for identical weights. Quantization and context limits differ, so check both columns before switching.
Benchmarks
Independent scores published alongside the catalogue.
- Coding index
- 44.2
About DeepSeek V3.2
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...
Specifications
| Model ID | deepseek/deepseek-v3.2 |
|---|---|
| Provider | DeepSeek |
| Context window | 164K tokens |
| Max output | 66K tokens |
| Input modalities | text |
| Output modalities | text |
| Knowledge cutoff | - |
| Open weights | Yes - deepseek-ai/DeepSeek-V3.2 |
| Released | December 1, 2025 |
Cheaper alternatives
Models that cost less than DeepSeek V3.2 while keeping at least half its context window and every input modality it supports.
Frequently asked
How much does DeepSeek: DeepSeek V3.2 cost?
$0.269 per million input tokens and $0.400 per million output tokens. Cached input reads cost $0.134 per million tokens.
What is the context window of DeepSeek: DeepSeek V3.2?
164K tokens, with up to 66K tokens of output per request.
Which provider serves DeepSeek: DeepSeek V3.2 cheapest?
GMICloud at $0.209 per million input tokens - 14× cheaper than SambaNova, the most expensive of the 15 hosts serving it.
Confirm against the source: DeepSeek official pricing.