DeepSeek-V3.2 Base: pricing, context window and benchmarks
DeepSeek-V3.2 Base is a multimodal model from DeepSeek. It offers a 128,000-token context window, costs $5.05 per million input tokens and $27.10 per million output tokens, and averages 80.2 across the index's benchmark suite.
Pricing
DeepSeek-V3.2 Base costs $5.05 per million input tokens and $27.10 per million output tokens. A blended 3:1 input-to-output mix works out to about $10.56 per million tokens. Cached input is listed at $0.56 per million tokens.
| Item | Price | Notes |
|---|---|---|
| Input | $5.05 | per 1M tokens |
| Output | $27.10 | per 1M tokens |
| Cached input | $0.56 | per 1M tokens |
| Blended 3:1 | $10.56 | estimate per 1M tokens |
| 1M in + 1M out | $32.15 | estimate |
Benchmarks
Scores below are modelled estimates on a 0–100 scale. Use them to compare relative position, not as official vendor results.
| Benchmark | Score | Rank | Percentile | Relative |
|---|---|---|---|---|
| MMLU-Pro | 84.3 | 74 | 88th | |
| GPQA Diamond | 77.1 | 103 | 83th | |
| SWE-bench Verified | 69.1 | 129 | 79th | |
| HumanEval | 82.8 | 163 | 73th | |
| MATH-500 | 83.7 | 133 | 78th | |
| AIME 2025 | 74.3 | 167 | 72th | |
| LiveCodeBench | 73.1 | 125 | 79th | |
| IFEval | 86.9 | 133 | 78th | |
| TruthfulQA | 83.5 | 92 | 85th | |
| HellaSwag | 90.2 | 79 | 87th | |
| MMMU | 76.7 | 118 | 80th | |
| Arena Elo | 1417.9 | 158 | 74th |
What DeepSeek-V3.2 Base is good for
DeepSeek-V3.2 Base is tuned toward coding work. It is a reasonable candidate for code generation, review, debugging and repository-scale tasks, especially when paired with a retrieval layer. At $5.05 per million input tokens, it sits in the mid-range of the index, and the 128,000-token context window is enough for large files and multi-file diffs.
Strengths
- Large 128,000-token context window for long documents and agent traces.
- Open weights make self-hosting, fine-tuning and data-residency review possible.
- Accepts more than text, including image input where the provider supports it.
- Coding benchmarks are a relative strength within its tier.
Trade-offs
- No standout weakness in the indexed specifications, but real-world behaviour still needs testing.
How it compares
| Model | Provider | Context | Input / 1M | Avg score |
|---|---|---|---|---|
| Command A Reasoning | Cohere | 400000 | $13.74 | 80.2 |
| DeepSeek-V3.2 Base | DeepSeek | 128000 | $5.05 | 80.2 |
| Doubao 1.5 Pro Pro | ByteDance | 128000 | $14.10 | 80.2 |
| GLM-4.6 | Zhipu AI | 128000 | $5.10 | 80.2 |
| Kimi K1.5 Chat | Moonshot AI | 128000 | $5.11 | 80.2 |
| MiniMax-M2 Chat | MiniMax | 200000 | $5.02 | 80.2 |
FAQ
How much does DeepSeek-V3.2 Base cost?
DeepSeek-V3.2 Base is listed at $5.05 per million input tokens and $27.10 per million output tokens. Cached input is $0.56 per million tokens. A blended 3:1 input-to-output mix is about $10.56 per million tokens.
What is the context window of DeepSeek-V3.2 Base?
DeepSeek-V3.2 Base has a context window of 128,000 tokens and can emit up to 128,000 output tokens per request in this index.
Is DeepSeek-V3.2 Base open weights?
Yes. DeepSeek-V3.2 Base is listed as an open-weight model under a open weights licence.
How does DeepSeek-V3.2 Base score on benchmarks?
DeepSeek-V3.2 Base averages 80.2 across the scored benchmarks in this index, ranking #113 of 600 models. Its strongest relative result is Arena Elo at 1417.9.
What is DeepSeek-V3.2 Base best for?
DeepSeek-V3.2 Base is tuned toward coding work. It is a reasonable candidate for code generation, review, debugging and repository-scale tasks, especially when paired with a retrieval layer. At $5.05 per million input tokens, it sits in the mid-range of the index, and the 128,000-token context window is enough for large files and multi-file diffs.