DeepSeek V3.2
- In Input modalities: Text Out Output modalities: Text
- Reasoning
- Reasoning can be off
- Tool use
- Open weights
Prices
6 offerings · prices as listeddeepseek-v3.2| Conditions/ 1M tokens | Input | Cache read | Output |
|---|---|---|---|
| Standard no longer listed since Oct 12, 2026 | $0.287 | $0.029 | $0.431 |
Prices from deepseek-v3.2 (opens in a new tab) (checked Oct 9, 2026)
deepseek-v3.2| Conditions/ 1M tokens | Input | Output |
|---|---|---|
| Standard no longer listed since Oct 11, 2026 | $0.57 | $1.71 |
Prices from Alibaba Cloud Model Studio pricing (opens in a new tab) (checked Oct 10, 2026)
deepseek-v3.2| Conditions/ 1M tokens | Input | Output |
|---|---|---|
| Standard no longer listed since Oct 11, 2026 | ¥2.00 | ¥3.00 |
Prices from 阿里云百炼：模型价格 (opens in a new tab) (checked Nov 2, 2026)
siliconflow/deepseek-v3.2| Conditions/ 1M tokens | Input | Output |
|---|---|---|
| Standard | ¥2.00 | ¥3.00 |
Prices from 阿里云百炼：模型价格 (opens in a new tab) (checked Nov 2, 2026)
Standard: the price a request with no conditions pays (the headline price elsewhere on the site).
Details
- Max thinking budget
- 32.77K
- Structured output
- —
- Knowledge cutoff
- Jul 2024
- Licence
- MIT License
- Family
- deepseek
- Model ID
Not stated on DeepSeek’s pages: max input, reasoning efforts, built-in tools, aliases.
Run locally
Weights: deepseek-ai/DeepSeek-V3.2- Parameters 685.4B
- Active 51.9B (estimated) · 256 experts, 8 per token
- Native precision FP8
- Layers 61
- KV cache per token 69 KB
- Config positions 160K (config.json max_position_embeddings)
- API context 125K (maker’s API)
- Needs
- ≈ 356.6 GB (weights 339.1 + KV cache 0.5 + overhead 17.0)
- Usable
- 21.6 GB (90% of its memory)
- Try
- it’s too large for this hardware; see the smallest that runs it below
| Precision | Weights | Needs at 8K | On your hardware | Longest context |
|---|---|---|---|---|
| FP8 (native)maker’s file | 642.1 GB | ≈ 674.8 | ✕Doesn’t fit | — |
| Q6_K | 524.2 GB | ≈ 550.9 | ✕Doesn’t fit | — |
| Q5_K_M | 454.0 GB | ≈ 477.2 | ✕Doesn’t fit | — |
| Q5_0 | 438.8 GB | ≈ 461.3 | ✕Doesn’t fit | — |
| Q4_K_M | 387.8 GB | ≈ 407.7 | ✕Doesn’t fit | — |
| Q4_0 | 359.0 GB | ≈ 377.5 | ✕Doesn’t fit | — |
| IQ4_XS | 339.1 GB | ≈ 356.6 | ✕Doesn’t fit | — |
| Q3_K_Mquality drops | 312.0 GB | ≈ 328.1 | ✕Doesn’t fit | — |
| IQ3_XXSquality drops | 244.1 GB | ≈ 256.9 | ✕Doesn’t fit | — |
| Q2_Kquality drops | 239.4 GB | ≈ 251.9 | ✕Doesn’t fit | — |
| IQ2_XXSquality drops | 164.4 GB | ≈ 173.1 | ✕Doesn’t fit | — |
Click a row to choose a precision. “Maker’s file” sizes are the maker’s own files; others are parameters × bits per weight.
Compared with other hardware
Each at the highest precision it runs. Click one for details.
Smallest that runs it (Q4_K_M, 8K): no single listed machine; about 6 × 80 GB GPUs.
Run commands
The maker publishes no files at this precision. Choose its native precision or one of its own builds, or quantise it yourself with llama.cpp.
Memory to fine-tune
Batch 1, 2,048-token sequences, gradient checkpointing; LoRA and QLoRA train rank-16 adapters on the attention projections.
Estimates; facts from the maker’s Hugging Face repositories, as of 2026-11-06. How it’s estimated
Cited scores
From third-party leaderboards, not measured by us- Epoch Capabilities Index 146.2
#75 of 118 listed models · CI 144.7–147.8
- LMArena Text 1,425
#79 of 163 listed models · CI 1,421–1,428 · 48,095 votes
- EQ-Bench Creative Writing v3 1,515
#51 of 93 listed models
- EQ-Bench Longform Writing 66.8
#45 of 92 listed models
- Epoch AI, 'Capabilities & benchmarking'. Published online at epoch.ai. Retrieved from 'https://epoch.ai/benchmarks'. Licensed CC BY 4.0. · License · as of Nov 5, 2026
- LMArena (Arena), leaderboard-dataset on Hugging Face, licensed CC BY 4.0 · License · as of Oct 28, 2026
- EQ-Bench by Samuel J. Paech (eqbench.com; EQ-bench/EQ-bench-site on GitHub), MIT License. · License · as of Oct 28, 2026
- changes we made
Included in coding plans
- Kiro Free Free50 credits per month
- Kiro Pro $20.00 / month1,000 credits per month
- Kiro Pro+ $40.00 / month2,000 credits per month
- Kiro Pro Max $100.00 / month5,000 credits per month
- Kiro Power $200.00 / month10,000 credits per month
Changes
All changes →- Now priced on Alibaba Cloud Model Studio, China Input ¥2.00 · Output ¥3.00
- Now priced on Alibaba Cloud Model Studio, International Input $0.57 · Output $1.71
Sources
- Text generation models - Qwen, DeepSeek, GLM | Model Studio - Alibaba Cloud Model Studio - Alibaba Cloud Documentation Center (opens in a new tab) docs.modelstudio.console.alibabacloud.com · checked Oct 10, 2026
- deepseek-v3.2 (opens in a new tab) alibabacloud.com · checked Oct 9, 2026
- Alibaba Cloud Model Studio pricing (opens in a new tab) alibabacloud.com · checked Oct 10, 2026
- DeepSeek (opens in a new tab) alibabacloud.com · checked Oct 9, 2026
- 硅基流动：模型价格in Chinese (opens in a new tab) siliconflow.cn · checked Nov 1, 2026
- 阿里云百炼：模型价格in Chinese (opens in a new tab) help.aliyun.com · checked Nov 2, 2026
- DeepSeek | DeepSeek V3.2 正式版：强化 Agent 能力，融入思考推理in Chinese (opens in a new tab) deepseek.com · checked Nov 5, 2026
- 模型上下架与更新-大模型服务平台百炼-阿里云_大模型服务平台百炼(Model Studio)-大模型服务平台百炼(Model Studio)-阿里云帮助中心in Chinese (opens in a new tab) help.aliyun.com · checked Nov 4, 2026
- Explore Qwen3, Wan2.6 & Latest Models on Model Studio - Alibaba Cloud Model Studio - Alibaba Cloud_Alibaba Cloud Model Studio(Model Studio) - Alibaba Cloud Model Studio - Alibaba Cloud Documentation Center (opens in a new tab) alibabacloud.com · checked Nov 3, 2026
- DeepSeek V3.2 Technical Documentation (opens in a new tab) fe-static.deepseek.com · checked Nov 6, 2026
- models.dev: deepseek/deepseek-v3.2 (community data, MIT) (opens in a new tab) github.com · checked Oct 15, 2026
- deepseek-ai/DeepSeek-V3.2 model card (opens in a new tab) huggingface.co · checked Nov 3, 2026
- deepseek-ai/DeepSeek-V3.2 config.json (opens in a new tab) huggingface.co · checked Nov 4, 2026
- deepseek-ai/DeepSeek-V3.2 files (opens in a new tab) huggingface.co · checked Nov 3, 2026
DeepSeek’s official policies: data use, privacy, terms →
Some values on this page are still waiting to be checked against official pages: a filled dot is verified, a hollow one is not yet (with or without a source on file). How we verify data
DeepSeek · China · deepseek.com
Related models
Snapshots, the same family and other models from this provider