DeepSeek-Coder: pricing, context window and benchmarks
DeepSeek-Coder is a coding model from DeepSeek. It offers a 32,768-token context window, costs $4.55 per million input tokens and $24.30 per million output tokens, and averages 78.1 across the index's benchmark suite.
Pricing
DeepSeek-Coder costs $4.55 per million input tokens and $24.30 per million output tokens. A blended 3:1 input-to-output mix works out to about $9.49 per million tokens. Cached input is listed at $0.48 per million tokens.
| Item | Price | Notes |
|---|---|---|
| Input | $4.55 | per 1M tokens |
| Output | $24.30 | per 1M tokens |
| Cached input | $0.48 | per 1M tokens |
| Blended 3:1 | $9.49 | estimate per 1M tokens |
| 1M in + 1M out | $28.85 | 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 | 78.4 | 229 | 62th | |
| GPQA Diamond | 74.0 | 167 | 72th | |
| SWE-bench Verified | 68.4 | 136 | 77th | |
| HumanEval | 81.5 | 195 | 68th | |
| MATH-500 | 79.9 | 209 | 65th | |
| AIME 2025 | 78.7 | 105 | 83th | |
| LiveCodeBench | 68.2 | 223 | 63th | |
| IFEval | 86.6 | 142 | 76th | |
| TruthfulQA | 81.5 | 140 | 77th | |
| HellaSwag | 87.3 | 162 | 73th | |
| MMMU | 74.1 | 183 | 70th | |
| Arena Elo | 1403.6 | 198 | 67th |
What DeepSeek-Coder is good for
DeepSeek-Coder 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 $4.55 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
- 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 | DeepSeek | 32768 | $4.55 | 78.1 |
| Gemini 3 Pro Mini | Google DeepMind | 128000 | $12.27 | 78.1 |
| Hunyuan-Large Chat | Tencent | 200000 | $4.35 | 78.1 |
| Qwen3-Coder Instruct | Alibaba Cloud | 128000 | $4.66 | 78.1 |
| ERNIE 4.0 Turbo | Baidu | 65536 | $12.00 | 78.0 |
| Llama 4 Scout Chat | Meta AI | 32768 | $4.66 | 78.0 |
FAQ
How much does DeepSeek-Coder cost?
DeepSeek-Coder is listed at $4.55 per million input tokens and $24.30 per million output tokens. Cached input is $0.48 per million tokens. A blended 3:1 input-to-output mix is about $9.49 per million tokens.
What is the context window of DeepSeek-Coder?
DeepSeek-Coder has a context window of 32,768 tokens and can emit up to 8,192 output tokens per request in this index.
Is DeepSeek-Coder open weights?
Yes. DeepSeek-Coder is listed as an open-weight model under a open weights licence.
How does DeepSeek-Coder score on benchmarks?
DeepSeek-Coder averages 78.1 across the scored benchmarks in this index, ranking #169 of 600 models. Its strongest relative result is Arena Elo at 1403.6.
What is DeepSeek-Coder best for?
DeepSeek-Coder 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 $4.55 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.