DeepSeek-Coder Base: pricing, context window and benchmarks
DeepSeek-Coder Base is a coding model from DeepSeek. It offers a 32,768-token context window, costs $3.53 per million input tokens and $12.53 per million output tokens, and averages 74.1 across the index's benchmark suite.
Pricing
DeepSeek-Coder Base costs $3.53 per million input tokens and $12.53 per million output tokens. A blended 3:1 input-to-output mix works out to about $5.78 per million tokens. Cached input is listed at $0.49 per million tokens.
| Item | Price | Notes |
|---|---|---|
| Input | $3.53 | per 1M tokens |
| Output | $12.53 | per 1M tokens |
| Cached input | $0.49 | per 1M tokens |
| Blended 3:1 | $5.78 | estimate per 1M tokens |
| 1M in + 1M out | $16.06 | 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 | 78.9 | 211 | 65th | |
| GPQA Diamond | 70.8 | 240 | 60th | |
| SWE-bench Verified | 59.3 | 337 | 44th | |
| HumanEval | 78.4 | 283 | 53th | |
| MATH-500 | 74.2 | 341 | 43th | |
| AIME 2025 | 71.7 | 223 | 63th | |
| LiveCodeBench | 67.3 | 242 | 60th | |
| IFEval | 81.6 | 282 | 53th | |
| TruthfulQA | 79.0 | 215 | 64th | |
| HellaSwag | 84.8 | 255 | 58th | |
| MMMU | 69.0 | 325 | 46th | |
| Arena Elo | 1388.5 | 255 | 58th |
What DeepSeek-Coder Base is good for
DeepSeek-Coder 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 $3.53 per million input tokens, it sits in the mid-range of the index, and the 32,768-token context window is enough for large files and multi-file diffs.
Strengths
- Input price of $3.53 per million tokens is at or below the index median.
- 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 |
|---|---|---|---|---|
| DeepSeek-Coder Base | DeepSeek | 32768 | $3.53 | 74.1 |
| GPT-5.1 Nano | OpenAI | 32768 | $9.56 | 74.1 |
| Gemma 3 Instruct | Google DeepMind | 128000 | $3.72 | 74.1 |
| Nova Lite Pro | Amazon | 200000 | $9.69 | 74.1 |
| QwQ Base | Alibaba Cloud | 65536 | $3.68 | 74.1 |
| Reka Flash | Reka AI | 128000 | $9.95 | 74.1 |
FAQ
How much does DeepSeek-Coder Base cost?
DeepSeek-Coder Base is listed at $3.53 per million input tokens and $12.53 per million output tokens. Cached input is $0.49 per million tokens. A blended 3:1 input-to-output mix is about $5.78 per million tokens.
What is the context window of DeepSeek-Coder Base?
DeepSeek-Coder Base has a context window of 32,768 tokens and can emit up to 32,768 output tokens per request in this index.
Is DeepSeek-Coder Base open weights?
Yes. DeepSeek-Coder Base is listed as an open-weight model under a open weights licence.
How does DeepSeek-Coder Base score on benchmarks?
DeepSeek-Coder Base averages 74.1 across the scored benchmarks in this index, ranking #272 of 600 models. Its strongest relative result is Arena Elo at 1388.5.
What is DeepSeek-Coder Base best for?
DeepSeek-Coder 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 $3.53 per million input tokens, it sits in the mid-range of the index, and the 32,768-token context window is enough for large files and multi-file diffs.