DeepSeek-V3.2 Instruct: pricing, context window and benchmarks
DeepSeek-V3.2 Instruct is a frontier model from DeepSeek. It offers a 128,000-token context window, costs $5.91 per million input tokens and $29.01 per million output tokens, and averages 83.7 across the index's benchmark suite.
Pricing
DeepSeek-V3.2 Instruct costs $5.91 per million input tokens and $29.01 per million output tokens. A blended 3:1 input-to-output mix works out to about $11.69 per million tokens. Cached input is listed at $0.55 per million tokens.
| Item | Price | Notes |
|---|---|---|
| Input | $5.91 | per 1M tokens |
| Output | $29.01 | per 1M tokens |
| Cached input | $0.55 | per 1M tokens |
| Blended 3:1 | $11.69 | estimate per 1M tokens |
| 1M in + 1M out | $34.92 | 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 | 84.6 | 68 | 89th | |
| GPQA Diamond | 80.9 | 44 | 93th | |
| SWE-bench Verified | 75.2 | 39 | 94th | |
| HumanEval | 90.2 | 42 | 93th | |
| MATH-500 | 86.3 | 87 | 86th | |
| AIME 2025 | 83.9 | 49 | 92th | |
| LiveCodeBench | 75.2 | 93 | 85th | |
| IFEval | 91.3 | 46 | 92th | |
| TruthfulQA | 84.3 | 78 | 87th | |
| HellaSwag | 88.8 | 116 | 81th | |
| MMMU | 80.4 | 53 | 91th | |
| Arena Elo | 1454.4 | 64 | 89th |
What DeepSeek-V3.2 Instruct is good for
DeepSeek-V3.2 Instruct sits in the frontier tier of the index. It is aimed at work where answer quality matters more than unit cost: hard reasoning, multi-step coding, document analysis and multimodal inputs. At $5.91 per million input tokens and a 128,000-token context window, it is a premium option, so reserve it for tasks where a cheaper model measurably fails.
Strengths
- Top-tier average benchmark score (83.7), placing it #66 of 600.
- 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.
- Reasoning-oriented profile, useful for multi-step problems and structured analysis.
Trade-offs
- Cost can climb quickly on long outputs and high-volume traffic.
How it compares
| Model | Provider | Context | Input / 1M | Avg score |
|---|---|---|---|---|
| Claude Haiku 4.5 Max | Anthropic | 200000 | $15.98 | 83.8 |
| Grok 3 Pro | xAI | 400000 | $16.61 | 83.8 |
| Claude 3.7 Sonnet Max | Anthropic | 128000 | $16.88 | 83.7 |
| Claude 3.7 Sonnet Reasoning | Anthropic | 128000 | $16.30 | 83.7 |
| DeepSeek-V3.2 Instruct | DeepSeek | 128000 | $5.91 | 83.7 |
| Gemini 3 Flash Reasoning | Google DeepMind | 128000 | $17.05 | 83.7 |
FAQ
How much does DeepSeek-V3.2 Instruct cost?
DeepSeek-V3.2 Instruct is listed at $5.91 per million input tokens and $29.01 per million output tokens. Cached input is $0.55 per million tokens. A blended 3:1 input-to-output mix is about $11.69 per million tokens.
What is the context window of DeepSeek-V3.2 Instruct?
DeepSeek-V3.2 Instruct has a context window of 128,000 tokens and can emit up to 4,096 output tokens per request in this index.
Is DeepSeek-V3.2 Instruct open weights?
Yes. DeepSeek-V3.2 Instruct is listed as an open-weight model under a open weights licence.
How does DeepSeek-V3.2 Instruct score on benchmarks?
DeepSeek-V3.2 Instruct averages 83.7 across the scored benchmarks in this index, ranking #66 of 600 models. Its strongest relative result is Arena Elo at 1454.4.
What is DeepSeek-V3.2 Instruct best for?
DeepSeek-V3.2 Instruct sits in the frontier tier of the index. It is aimed at work where answer quality matters more than unit cost: hard reasoning, multi-step coding, document analysis and multimodal inputs. At $5.91 per million input tokens and a 128,000-token context window, it is a premium option, so reserve it for tasks where a cheaper model measurably fails.