# DeepSeek-V3.2 by DeepSeek: Open-Weight Model Source: https://savrn.com/models/deepseek-v3-2 Markdown alternate of the page above; the site index is https://savrn.com/llms.txt --- ## Runs On What it takes to serve DeepSeek-V3.2 (685.4B parameters): the memory its weights need at each precision, and the cheapest way to rent enough data-center GPUs to hold them. | Precision | Weights | Memory needed | Cheapest setup | Per hour | Also fits | | --- | --- | --- | --- | --- | --- | | 16-bit | 1370.7 GB | 1644.9 GB | 7x [MI325X](https://savrn.com/ai-index/pricing/gpus/mi325x) (256 GB) Vultr | $14.00 | [6x MI355X](https://savrn.com/ai-index/pricing/gpus/mi355x) $15.54 · [7x B300](https://savrn.com/ai-index/pricing/gpus/b300) $46.20 | | 8-bit | 685.4 GB | 822.4 GB | 3x [MI355X](https://savrn.com/ai-index/pricing/gpus/mi355x) (288 GB) Vultr | $7.77 | [4x MI325X](https://savrn.com/ai-index/pricing/gpus/mi325x) $8.00 · [5x MI300X](https://savrn.com/ai-index/pricing/gpus/mi300x) $9.25 | | 4-bit | 342.7 GB | 411.2 GB | 2x [MI325X](https://savrn.com/ai-index/pricing/gpus/mi325x) (256 GB) Vultr | $4.00 | [2x MI355X](https://savrn.com/ai-index/pricing/gpus/mi355x) $5.18 · [3x MI300X](https://savrn.com/ai-index/pricing/gpus/mi300x) $5.55 | Memory is the weights at that precision plus 20% for the runtime and a short context; a long context needs more. Prices are the lowest on-demand hourly rates in the [SAVRN Index](https://savrn.com/ai-index/pricing/gpus), read Oct 23, 2026. [DeepSeek-V3.2 on every accelerator the SAVRN Index prices, at every precision](https://savrn.com/models/deepseek-v3-2/gpus) ## SAVRN's Notes on DeepSeek-V3.2 685.4B parameters and 3.3M downloads a month make this the heavyweight of the group, and the Index prices a three-rung ladder: 4-bit needs 411.2 GB, two MI325X at $4.00 an hour; 8-bit needs 822.4 GB, three MI355X at $7.77; 16-bit needs 1,644.9 GB, seven MI325X at $14.00. The mixture-of-experts design, DeepSeek Sparse Attention and a 163,840-token window are aimed at reasoning and agent work, and the safetensors download alone is 689.5 GB across 192 files. Rent-or-own is a real choice: DeepInfra lists $0.26 in and $0.38 out per million tokens, Novita $0.27 and $0.40 — price your monthly volume against the $4.00-an-hour 4-bit floor before buying cards. MIT is the most permissive license on this list, commercial use with just the notice preserved. Treat the recorded evals (Claw-Eval 42.2 general and 31.6 multi-turn, WildClawBench overall 34, all third-party reported) as agentic-suite indicators, not general capability scores. SAVRN Research ## Model Card By DeepSeek, published under mit, revision a7e62ac04ecb. ### DeepSeek-V3.2: Efficient Reasoning & Agentic AI [Technical Report](https://huggingface.co/deepseek-ai/DeepSeek-V3.2/blob/main/assets/paper.pdf) ### Introduction We introduce DeepSeek-V3.2, a model that harmonizes high computational efficiency with superior reasoning and agent performance. Our approach is built upon three key technical breakthroughs: 1. DeepSeek Sparse Attention (DSA): We introduce DSA, an efficient attention mechanism that substantially reduces computational complexity while preserving model performance, specifically optimized for long-context scenarios. 2. Scalable Reinforcement Learning Framework: By implementing a robust RL protocol and scaling post-training compute, DeepSeek-V3.2 performs comparably to GPT-5. Notably, our high-compute variant, DeepSeek-V3.2-Speciale, surpasses GPT-5 and exhibits reasoning proficiency on par with Gemini-3.0-Pro. - Achievement:Gold-medal performance in the 2025 International Mathematical Olympiad (IMO) and International Olympiad in Informatics (IOI). 3. Large-Scale Agentic Task Synthesis Pipeline: To integrate reasoning into tool-use scenarios, we developed a novel synthesis pipeline that systematically generates training data at scale. This facilitates scalable agentic post-training, improving compliance and generalization in complex interactive environments. [Read the full model card (572 words)](https://savrn.com/models/deepseek-v3-2/card) ## Configuration Architecture DeepseekV32ForCausalLM Context length (tokens) 163,840 Layers 61 Hidden size 7,168 Feed-forward size 18,432 Attention heads 128 Key/value heads 128 Vocabulary size 129,280 Routed experts 256 Experts active per token 8 RoPE base 10,000 Stored precision bfloat16 Model type deepseek_v32 Quantization fp8 ## Identity and Version Repository deepseek-ai/DeepSeek-V3.2 Publisher DeepSeek Task Text generation Modality Text Library transformers Parameters 685.4B parameters Languages Not stated by the source Revision a7e62ac04ecb2c0a54d736dc46601c5606cf10a6 First published 2025-12-01 Last updated 2025-12-01 ## Files and Weights 192 files, 689.5 GB in total. The weights are 163 files totalling 689.5 GB in safetensors. Weights163 files · 689.5 GB Configuration15 files · 15.4 MB Tokenizer2 files · 7.8 MB Documentation3 files · 9.0 KB Other8 files · 1.8 MB Repository1 file · 1.6 KB Every file | File | Type | Size | SHA-256 | | --- | --- | --- | --- | | model-00001-of-000163.safetensors | Weights | 5.2 GB | a20d4376cb0f | | model-00002-of-000163.safetensors | Weights | 4.3 GB | ca4cbcfcfbe0 | | model-00003-of-000163.safetensors | Weights | 4.3 GB | 9addd18a0a46 | | model-00004-of-000163.safetensors | Weights | 4.3 GB | ce280c84088e | | model-00005-of-000163.safetensors | Weights | 4.3 GB | 0aeaa31d376f | | model-00006-of-000163.safetensors | Weights | 4.3 GB | f2caca0cf47e | | model-00007-of-000163.safetensors | Weights | 4.3 GB | 3e8c95f758f7 | | model-00008-of-000163.safetensors | Weights | 4.3 GB | e588199641ea | | model-00009-of-000163.safetensors | Weights | 4.3 GB | 7d3cec332f93 | | model-00010-of-000163.safetensors | Weights | 4.3 GB | 36c8ee43fba4 | | model-00011-of-000163.safetensors | Weights | 4.3 GB | 3092be80417e | | model-00012-of-000163.safetensors | Weights | 1.5 GB | b04e90caba83 | | model-00013-of-000163.safetensors | Weights | 4.3 GB | abbe4de59b34 | | model-00014-of-000163.safetensors | Weights | 4.3 GB | 51f32772258f | | model-00015-of-000163.safetensors | Weights | 4.3 GB | f88eda8849a7 | | model-00016-of-000163.safetensors | Weights | 4.3 GB | 877d653ea176 | | model-00017-of-000163.safetensors | Weights | 4.3 GB | d33bbd747e7b | | model-00018-of-000163.safetensors | Weights | 4.3 GB | 71e34312df5c | | model-00019-of-000163.safetensors | Weights | 4.3 GB | f0764e374a4e | | model-00020-of-000163.safetensors | Weights | 4.3 GB | 5847909aca3a | | model-00021-of-000163.safetensors | Weights | 4.3 GB | 42412fde8f98 | | model-00022-of-000163.safetensors | Weights | 4.3 GB | de909ed548c1 | | model-00023-of-000163.safetensors | Weights | 4.3 GB | 6fd845c676f5 | | model-00024-of-000163.safetensors | Weights | 4.3 GB | 3743fa6abc0f | | model-00025-of-000163.safetensors | Weights | 4.3 GB | 167fe31e96d8 | | model-00026-of-000163.safetensors | Weights | 4.3 GB | 244af9059f19 | | model-00027-of-000163.safetensors | Weights | 4.3 GB | e5ba88b98db2 | | model-00028-of-000163.safetensors | Weights | 4.3 GB | 08d2d16b9b70 | | model-00029-of-000163.safetensors | Weights | 4.3 GB | 26490465ffb6 | | model-00030-of-000163.safetensors | Weights | 4.3 GB | f9a4a75bcd87 | | model-00031-of-000163.safetensors | Weights | 4.3 GB | 18e2144cb7ce | | model-00032-of-000163.safetensors | Weights | 4.3 GB | c11e401edfac | | model-00033-of-000163.safetensors | Weights | 4.3 GB | 5023f78d52f4 | | model-00034-of-000163.safetensors | Weights | 1.9 GB | 25d0de971c7a | | model-00035-of-000163.safetensors | Weights | 4.3 GB | 6d9ca99e5a62 | | model-00036-of-000163.safetensors | Weights | 4.3 GB | 8fd1c8f7856d | | model-00037-of-000163.safetensors | Weights | 4.3 GB | 018aa688bab6 | | model-00038-of-000163.safetensors | Weights | 4.3 GB | f65f19202f42 | | model-00039-of-000163.safetensors | Weights | 4.3 GB | 284781696e8d | | model-00040-of-000163.safetensors | Weights | 4.3 GB | 589352dbfa33 | | model-00041-of-000163.safetensors | Weights | 4.3 GB | 435ad4d3b468 | | model-00042-of-000163.safetensors | Weights | 4.3 GB | 0ee57f8fa923 | | model-00043-of-000163.safetensors | Weights | 4.3 GB | 8a14306fc77a | | model-00044-of-000163.safetensors | Weights | 4.3 GB | 14127198dbed | | model-00045-of-000163.safetensors | Weights | 4.3 GB | d3b75d7a61fc | | model-00046-of-000163.safetensors | Weights | 4.3 GB | 2137ebf1537f | | model-00047-of-000163.safetensors | Weights | 4.3 GB | 71a16ee2120f | | model-00048-of-000163.safetensors | Weights | 4.3 GB | ce759a29e97a | | model-00049-of-000163.safetensors | Weights | 4.3 GB | 9f4f9db35940 | | model-00050-of-000163.safetensors | Weights | 4.3 GB | 052f6cf94229 | | model-00051-of-000163.safetensors | Weights | 4.3 GB | f4affbf46f9c | | model-00052-of-000163.safetensors | Weights | 4.3 GB | 59ec824f83d1 | | model-00053-of-000163.safetensors | Weights | 4.3 GB | 80b44c76d9a0 | | model-00054-of-000163.safetensors | Weights | 4.3 GB | 2940501b0551 | | model-00055-of-000163.safetensors | Weights | 4.3 GB | 5db89a321814 | | model-00056-of-000163.safetensors | Weights | 1.9 GB | 7c291b4057b0 | | model-00057-of-000163.safetensors | Weights | 4.3 GB | bfee7a8078cb | | model-00058-of-000163.safetensors | Weights | 4.3 GB | 61826eb57e15 | | model-00059-of-000163.safetensors | Weights | 4.3 GB | a61c1fa7e19a | | model-00060-of-000163.safetensors | Weights | 4.3 GB | e484ef9d1e7e | | model-00061-of-000163.safetensors | Weights | 4.3 GB | 442638b1c9d9 | | model-00062-of-000163.safetensors | Weights | 4.3 GB | f6cd3e2ff6ca | | model-00063-of-000163.safetensors | Weights | 4.3 GB | a43d4644ab3c | | model-00064-of-000163.safetensors | Weights | 4.3 GB | 982c6d763d0f | | model-00065-of-000163.safetensors | Weights | 4.3 GB | 729e37c6379f | | model-00066-of-000163.safetensors | Weights | 4.3 GB | c5192ad78c7f | | model-00067-of-000163.safetensors | Weights | 4.3 GB | 4d8f176bdd3c | | model-00068-of-000163.safetensors | Weights | 4.3 GB | bcef3c24fc0e | | model-00069-of-000163.safetensors | Weights | 4.3 GB | d1222e0940cd | | model-00070-of-000163.safetensors | Weights | 4.3 GB | f21c317c4d5e | | model-00071-of-000163.safetensors | Weights | 4.3 GB | e7ff81c4f916 | | model-00072-of-000163.safetensors | Weights | 4.3 GB | 4b537a0a4ae3 | | model-00073-of-000163.safetensors | Weights | 4.3 GB | 95c2a86af5da | | model-00074-of-000163.safetensors | Weights | 4.3 GB | 891648dde77c | | model-00075-of-000163.safetensors | Weights | 4.3 GB | d117d7f3ca33 | | model-00076-of-000163.safetensors | Weights | 4.3 GB | 8da7550276e8 | | model-00077-of-000163.safetensors | Weights | 4.3 GB | 359e893bda3a | | model-00078-of-000163.safetensors | Weights | 1.9 GB | de731d86997a | | model-00079-of-000163.safetensors | Weights | 4.3 GB | d79b8523cb68 | | model-00080-of-000163.safetensors | Weights | 4.3 GB | 7d88f3162146 | | model-00081-of-000163.safetensors | Weights | 4.3 GB | 677b816994f1 | | model-00082-of-000163.safetensors | Weights | 4.3 GB | 1428240eb9f6 | | model-00083-of-000163.safetensors | Weights | 4.3 GB | 3256b3288474 | | model-00084-of-000163.safetensors | Weights | 4.3 GB | 8f18fda3b08c | | model-00085-of-000163.safetensors | Weights | 4.3 GB | 2b98cf90e4c7 | | model-00086-of-000163.safetensors | Weights | 4.3 GB | ded19defd6fb | | model-00087-of-000163.safetensors | Weights | 4.3 GB | 2cd53d5c4212 | | model-00088-of-000163.safetensors | Weights | 4.3 GB | 4de9e378923f | | model-00089-of-000163.safetensors | Weights | 4.3 GB | 3c8ea189d66d | | model-00090-of-000163.safetensors | Weights | 4.3 GB | ad0cd7d5bcd6 | | model-00091-of-000163.safetensors | Weights | 4.3 GB | 183fe5e20893 | | model-00092-of-000163.safetensors | Weights | 4.3 GB | 0db72c733857 | | model-00093-of-000163.safetensors | Weights | 4.3 GB | ba824a8f5e50 | | model-00094-of-000163.safetensors | Weights | 4.3 GB | 596ab0761f3a | | model-00095-of-000163.safetensors | Weights | 4.3 GB | 0bece07ecd57 | | model-00096-of-000163.safetensors | Weights | 4.3 GB | f99f30cd2a04 | | model-00097-of-000163.safetensors | Weights | 4.3 GB | 5c0ad05dce5a | | model-00098-of-000163.safetensors | Weights | 4.3 GB | aeefc1152d32 | | model-00099-of-000163.safetensors | Weights | 4.3 GB | 93cacdf1c358 | | model-00100-of-000163.safetensors | Weights | 1.9 GB | ffeddd504a54 | | model-00101-of-000163.safetensors | Weights | 4.3 GB | 278020c3a0a0 | | model-00102-of-000163.safetensors | Weights | 4.3 GB | ffd588fd7aaa | | model-00103-of-000163.safetensors | Weights | 4.3 GB | 3acb6068a0f4 | | model-00104-of-000163.safetensors | Weights | 4.3 GB | a7ee6148cc85 | | model-00105-of-000163.safetensors | Weights | 4.3 GB | 88dd1665e59e | | model-00106-of-000163.safetensors | Weights | 4.3 GB | cec388cd8f9d | | model-00107-of-000163.safetensors | Weights | 4.3 GB | 897f8e2dd87c | | model-00108-of-000163.safetensors | Weights | 4.3 GB | 2a25ec05acd2 | | model-00109-of-000163.safetensors | Weights | 4.3 GB | 8be69408eacb | | model-00110-of-000163.safetensors | Weights | 4.3 GB | a65a239cc05d | | model-00111-of-000163.safetensors | Weights | 4.3 GB | 3e483a7229f0 | | model-00112-of-000163.safetensors | Weights | 4.3 GB | 64cf509f9304 | | model-00113-of-000163.safetensors | Weights | 4.3 GB | 5f07054b0b12 | | model-00114-of-000163.safetensors | Weights | 4.3 GB | c5f9b51692a5 | | model-00115-of-000163.safetensors | Weights | 4.3 GB | cb5b8951313b | | model-00116-of-000163.safetensors | Weights | 4.3 GB | 0e059ed5c223 | | model-00117-of-000163.safetensors | Weights | 4.3 GB | 9eb47e82aeae | | model-00118-of-000163.safetensors | Weights | 4.3 GB | 8667166977c0 | | model-00119-of-000163.safetensors | Weights | 4.3 GB | cd011f63a9d8 | | model-00120-of-000163.safetensors | Weights | 4.3 GB | d3e84b88c644 | | model-00121-of-000163.safetensors | Weights | 4.3 GB | de00ff563eae | | model-00122-of-000163.safetensors | Weights | 1.9 GB | d7db77a4dd01 | | model-00123-of-000163.safetensors | Weights | 4.3 GB | c0fc1ff16dcf | | model-00124-of-000163.safetensors | Weights | 4.3 GB | 59ecab8be5b9 | | model-00125-of-000163.safetensors | Weights | 4.3 GB | 38a59f70eb21 | | model-00126-of-000163.safetensors | Weights | 4.3 GB | 021a8c97b7ab | | model-00127-of-000163.safetensors | Weights | 4.3 GB | 416271db6327 | | model-00128-of-000163.safetensors | Weights | 4.3 GB | 81f47704507f | | model-00129-of-000163.safetensors | Weights | 4.3 GB | 9bae27016f20 | | model-00130-of-000163.safetensors | Weights | 4.3 GB | d16fefaead36 | | model-00131-of-000163.safetensors | Weights | 4.3 GB | e7ea72470022 | | model-00132-of-000163.safetensors | Weights | 4.3 GB | ca13cf0c562a | | model-00133-of-000163.safetensors | Weights | 4.3 GB | c3c419e47573 | | model-00134-of-000163.safetensors | Weights | 4.3 GB | b0dd13676ced | | model-00135-of-000163.safetensors | Weights | 4.3 GB | b6979bb49b28 | | model-00136-of-000163.safetensors | Weights | 4.3 GB | f43ec1b90698 | | model-00137-of-000163.safetensors | Weights | 4.3 GB | 7e4a0dc5e67a | | model-00138-of-000163.safetensors | Weights | 4.3 GB | 74769133c5f5 | | model-00139-of-000163.safetensors | Weights | 4.3 GB | 582e2f9d7124 | | model-00140-of-000163.safetensors | Weights | 4.3 GB | a07d2c16ad9b | | model-00141-of-000163.safetensors | Weights | 3.2 GB | 267a3bd6abcc | | model-00142-of-000163.safetensors | Weights | 4.3 GB | 00290593119b | | model-00143-of-000163.safetensors | Weights | 4.3 GB | 54717d9742a5 | | model-00144-of-000163.safetensors | Weights | 4.3 GB | d02691e8c9fb | | model-00145-of-000163.safetensors | Weights | 4.3 GB | 063580b98834 | | model-00146-of-000163.safetensors | Weights | 4.3 GB | 062e0dc7bcab | | model-00147-of-000163.safetensors | Weights | 4.3 GB | 18ce8926102d | | model-00148-of-000163.safetensors | Weights | 4.3 GB | 42dec5e5e2c6 | | model-00149-of-000163.safetensors | Weights | 4.3 GB | 5619ea0af4a7 | | model-00150-of-000163.safetensors | Weights | 4.3 GB | c43edb1641b8 | | model-00151-of-000163.safetensors | Weights | 4.3 GB | 8420a702f53e | | model-00152-of-000163.safetensors | Weights | 4.3 GB | 2132b8c21e1b | | model-00153-of-000163.safetensors | Weights | 4.3 GB | ab87c6c695bf | | model-00154-of-000163.safetensors | Weights | 4.3 GB | d6968de6684a | | model-00155-of-000163.safetensors | Weights | 4.3 GB | d54400800788 | | model-00156-of-000163.safetensors | Weights | 4.3 GB | 2e9b3ffac4fc | | model-00157-of-000163.safetensors | Weights | 4.3 GB | 24d1fafd6e13 | | model-00158-of-000163.safetensors | Weights | 4.3 GB | 95f1b28e15ec | | model-00159-of-000163.safetensors | Weights | 4.3 GB | 08cfd9dc2a2c | | model-00160-of-000163.safetensors | Weights | 5.3 GB | c0e810a252ae | | model-00161-of-000163.safetensors | Weights | 4.3 GB | a3bdc2e1a96d | | model-00162-of-000163.safetensors | Weights | 4.3 GB | 96ec087d8534 | | model-00163-of-000163.safetensors | Weights | 6.6 GB | dda8c6d06366 | | assets/olympiad_cases/ioi_submissions_final.json | Configuration | 2.6 MB | — | | assets/olympiad_cases/wf_submissions.json | Configuration | 3.5 MB | — | | config.json | Configuration | 1.6 KB | — | | encoding/encoding_dsv32.py | Configuration | 14.3 KB | — | | encoding/test_encoding_dsv32.py | Configuration | 2.1 KB | — | | encoding/test_input.json | Configuration | 6.7 KB | — | | encoding/test_input_search_w_date.json | Configuration | 183.8 KB | — | | encoding/test_input_search_wo_date.json | Configuration | 84.2 KB | — | | generation_config.json | Configuration | 171 B | — | | inference/config_671B_v3.2.json | Configuration | 605 B | — | | inference/convert.py | Configuration | 3.9 KB | — | | inference/generate.py | Configuration | 7.8 KB | — | | inference/kernel.py | Configuration | 10.0 KB | — | | inference/model.py | Configuration | 38.6 KB | — | | model.safetensors.index.json | Configuration | 8.9 MB | — | | LICENSE | Documentation | 1.1 KB | — | | README.md | Documentation | 7.4 KB | — | | inference/README.md | Documentation | 548 B | — | | assets/benchmark.png | Other | 466.8 KB | 08a78f9d224d | | assets/olympiad_cases/CMO2025.jsonl | Other | 79.8 KB | — | | assets/olympiad_cases/IMO2025.jsonl | Other | 63.0 KB | — | | assets/paper.pdf | Other | 907.1 KB | f6fda5753db7 | | encoding/test_output.txt | Other | 5.1 KB | — | | encoding/test_output_search_w_date.txt | Other | 172.5 KB | — | | encoding/test_output_search_wo_date.txt | Other | 76.6 KB | — | | inference/requirements.txt | Other | 70 B | — | | .gitattributes | Repository | 1.6 KB | — | | tokenizer.json | Tokenizer | 7.8 MB | — | | tokenizer_config.json | Tokenizer | 795 B | — | ## License and Download License mit Access Open weights, no gate Download size 689.5 GB [Download from DeepSeek](https://huggingface.co/deepseek-ai/DeepSeek-V3.2) Released by DeepSeek through its official repository on Hugging Face. [Read the license](https://opensource.org/license/mit). ## Built From - Derived from deepseek-ai/DeepSeek-V3.2-Exp-Base ## Evaluations Each result is shown as reported, with the conditions its reporter stated. None is a SAVRN measurement. A comparison lines two results up only when their configuration, unit and setup are all stated and identical. | Benchmark | Conditions | Result | Reported by | Revision | Date | | --- | --- | --- | --- | --- | --- | | claw-eval/Claw-Eval | Task generalMetric generalSetup Pass³% \| N=3 \| 161 tasksComparison conditions not established | 42.2 | Claw-Eval Leaderboard Reported by a third party | Evaluated revision not stated | 2026-04-23 | | claw-eval/Claw-Eval | Task multi_turnMetric multi_turnSetup Pass³% \| N=3 \| 38 tasksComparison conditions not established | 31.6 | Claw-Eval Leaderboard Reported by a third party | Evaluated revision not stated | 2026-04-23 | | internlm/WildClawBench | Task avg_costMetric avg_costComparison conditions not established | 11.4 | WildClawBench Reported by a third party | Evaluated revision not stated | 2026-05-22 | | internlm/WildClawBench | Task avg_timeMetric avg_timeComparison conditions not established | 549 | WildClawBench Reported by a third party | Evaluated revision not stated | 2026-05-22 | | internlm/WildClawBench | Task overallMetric overallComparison conditions not established | 34 | WildClawBench Reported by a third party | Evaluated revision not stated | 2026-05-22 | | open-agent-leaderboard/results | Task appworldMetric appworldSetup agent: React + ShortlistingComparison conditions not established | 0.04 | Open Agent Leaderboard Reported by a third party | Evaluated revision not stated | 2026-05-18 | | open-agent-leaderboard/results | Task appworldMetric appworldSetup agent: OpenAI SoloComparison conditions not established | 0.06 | Open Agent Leaderboard Reported by a third party | Evaluated revision not stated | 2026-05-18 | | open-agent-leaderboard/results | Task appworldMetric appworldSetup agent: SmolagentComparison conditions not established | 0.13 | Open Agent Leaderboard Reported by a third party | Evaluated revision not stated | 2026-05-18 | | open-agent-leaderboard/results | Task appworldMetric appworldSetup agent: Claude CodeComparison conditions not established | 0.03 | Open Agent Leaderboard Reported by a third party | Evaluated revision not stated | 2026-05-18 | | open-agent-leaderboard/results | Task appworldMetric appworldSetup agent: ReactComparison conditions not established | 0.09 | Open Agent Leaderboard Reported by a third party | Evaluated revision not stated | 2026-05-18 | | open-agent-leaderboard/results | Task browsecomp_plusMetric browsecomp_plusSetup agent: SmolagentComparison conditions not established | 0.21 | Open Agent Leaderboard Reported by a third party | Evaluated revision not stated | 2026-05-18 | | open-agent-leaderboard/results | Task browsecomp_plusMetric browsecomp_plusSetup agent: OpenAI SoloComparison conditions not established | 0.3 | Open Agent Leaderboard Reported by a third party | Evaluated revision not stated | 2026-05-18 | | open-agent-leaderboard/results | Task browsecomp_plusMetric browsecomp_plusSetup agent: Claude CodeComparison conditions not established | 0.48 | Open Agent Leaderboard Reported by a third party | Evaluated revision not stated | 2026-05-18 | | open-agent-leaderboard/results | Task browsecomp_plusMetric browsecomp_plusSetup agent: React + ShortlistingComparison conditions not established | 0.36 | Open Agent Leaderboard Reported by a third party | Evaluated revision not stated | 2026-05-18 | | open-agent-leaderboard/results | Task browsecomp_plusMetric browsecomp_plusSetup agent: ReactComparison conditions not established | 0.36 | Open Agent Leaderboard Reported by a third party | Evaluated revision not stated | 2026-05-18 | | open-agent-leaderboard/results | Task overallMetric overallSetup agent: ReactComparison conditions not established | 0.4585 | Open Agent Leaderboard Reported by a third party | Evaluated revision not stated | 2026-05-18 | | open-agent-leaderboard/results | Task overallMetric overallSetup agent: React + ShortlistingComparison conditions not established | 0.446 | Open Agent Leaderboard Reported by a third party | Evaluated revision not stated | 2026-05-18 | | open-agent-leaderboard/results | Task overallMetric overallSetup agent: Claude CodeComparison conditions not established | 0.4158 | Open Agent Leaderboard Reported by a third party | Evaluated revision not stated | 2026-05-18 | | open-agent-leaderboard/results | Task overallMetric overallSetup agent: OpenAI SoloComparison conditions not established | 0.3217 | Open Agent Leaderboard Reported by a third party | Evaluated revision not stated | 2026-05-18 | | open-agent-leaderboard/results | Task overallMetric overallSetup agent: SmolagentComparison conditions not established | 0.4092 | Open Agent Leaderboard Reported by a third party | Evaluated revision not stated | 2026-05-18 | | open-agent-leaderboard/results | Task swebenchMetric swebenchSetup agent: OpenAI SoloComparison conditions not established | 0.7368 | Open Agent Leaderboard Reported by a third party | Evaluated revision not stated | 2026-05-18 | | open-agent-leaderboard/results | Task swebenchMetric swebenchSetup agent: SmolagentComparison conditions not established | 0.56 | Open Agent Leaderboard Reported by a third party | Evaluated revision not stated | 2026-05-18 | | open-agent-leaderboard/results | Task swebenchMetric swebenchSetup agent: React + ShortlistingComparison conditions not established | 0.6875 | Open Agent Leaderboard Reported by a third party | Evaluated revision not stated | 2026-05-18 | | open-agent-leaderboard/results | Task swebenchMetric swebenchSetup agent: ReactComparison conditions not established | 0.6875 | Open Agent Leaderboard Reported by a third party | Evaluated revision not stated | 2026-05-18 | | open-agent-leaderboard/results | Task swebenchMetric swebenchSetup agent: Claude CodeComparison conditions not established | 0.64 | Open Agent Leaderboard Reported by a third party | Evaluated revision not stated | 2026-05-18 | | open-agent-leaderboard/results | Task taubench_airlineMetric taubench_airlineSetup agent: React + ShortlistingComparison conditions not established | 0.56 | Open Agent Leaderboard Reported by a third party | Evaluated revision not stated | 2026-05-18 | | open-agent-leaderboard/results | Task taubench_airlineMetric taubench_airlineSetup agent: SmolagentComparison conditions not established | 0.6 | Open Agent Leaderboard Reported by a third party | Evaluated revision not stated | 2026-05-18 | | open-agent-leaderboard/results | Task taubench_airlineMetric taubench_airlineSetup agent: ReactComparison conditions not established | 0.56 | Open Agent Leaderboard Reported by a third party | Evaluated revision not stated | 2026-05-18 | | open-agent-leaderboard/results | Task taubench_airlineMetric taubench_airlineSetup agent: OpenAI SoloComparison conditions not established | 0.2 | Open Agent Leaderboard Reported by a third party | Evaluated revision not stated | 2026-05-18 | | open-agent-leaderboard/results | Task taubench_airlineMetric taubench_airlineSetup agent: Claude CodeComparison conditions not established | 0.28 | Open Agent Leaderboard Reported by a third party | Evaluated revision not stated | 2026-05-18 | | open-agent-leaderboard/results | Task taubench_retailMetric taubench_retailSetup agent: SmolagentComparison conditions not established | 0.77 | Open Agent Leaderboard Reported by a third party | Evaluated revision not stated | 2026-05-18 | | open-agent-leaderboard/results | Task taubench_retailMetric taubench_retailSetup agent: Claude CodeComparison conditions not established | 0.65 | Open Agent Leaderboard Reported by a third party | Evaluated revision not stated | 2026-05-18 | | open-agent-leaderboard/results | Task taubench_retailMetric taubench_retailSetup agent: React + ShortlistingComparison conditions not established | 0.82 | Open Agent Leaderboard Reported by a third party | Evaluated revision not stated | 2026-05-18 | | open-agent-leaderboard/results | Task taubench_retailMetric taubench_retailSetup agent: ReactComparison conditions not established | 0.82 | Open Agent Leaderboard Reported by a third party | Evaluated revision not stated | 2026-05-18 | | open-agent-leaderboard/results | Task taubench_retailMetric taubench_retailSetup agent: OpenAI SoloComparison conditions not established | 0.19 | Open Agent Leaderboard Reported by a third party | Evaluated revision not stated | 2026-05-18 | | open-agent-leaderboard/results | Task taubench_telecomMetric taubench_telecomSetup agent: Claude CodeComparison conditions not established | 0.61 | Open Agent Leaderboard Reported by a third party | Evaluated revision not stated | 2026-05-18 | | open-agent-leaderboard/results | Task taubench_telecomMetric taubench_telecomSetup agent: React + ShortlistingComparison conditions not established | 0.71 | Open Agent Leaderboard Reported by a third party | Evaluated revision not stated | 2026-05-18 | | open-agent-leaderboard/results | Task taubench_telecomMetric taubench_telecomSetup agent: SmolagentComparison conditions not established | 0.84 | Open Agent Leaderboard Reported by a third party | Evaluated revision not stated | 2026-05-18 | | open-agent-leaderboard/results | Task taubench_telecomMetric taubench_telecomSetup agent: ReactComparison conditions not established | 0.71 | Open Agent Leaderboard Reported by a third party | Evaluated revision not stated | 2026-05-18 | | open-agent-leaderboard/results | Task taubench_telecomMetric taubench_telecomSetup agent: OpenAI SoloComparison conditions not established | 0.18 | Open Agent Leaderboard Reported by a third party | Evaluated revision not stated | 2026-05-18 | ## Memory Requirements | Precision | Weights in memory | | --- | --- | | As published | 689.5 GB | | 16-bit | 1370.7 GB | | 8-bit | 685.4 GB | | 4-bit | 342.7 GB | Weights only, from the published parameter count; the key-value cache and runtime add to this. ## Hosted Prices | Host | Input / output | Unit | Observed | | --- | --- | --- | --- | | DeepInfra | $0.26 / $0.38 | input / output, per million tokens | Oct 22, 2026 | | Novita | $0.27 / $0.40 | input / output, per million tokens | Sep 28, 2026 | From the [SAVRN Index](https://savrn.com/ai-index/pricing/open-models). ## Compare DeepSeek-V3.2 - [DeepSeek-V3.2 vs GLM-5.2](https://savrn.com/models/compare/deepseek-v3-2-vs-glm-5-2) - [DeepSeek-R1 vs DeepSeek-V3.2](https://savrn.com/models/compare/deepseek-r1-vs-deepseek-v3-2) - [DeepSeek-V3-0324 vs DeepSeek-V3.2](https://savrn.com/models/compare/deepseek-v3-0324-vs-deepseek-v3-2) ## Questions About DeepSeek-V3.2 ### How much GPU memory does DeepSeek-V3.2 need? About 1644.9 GB at 16-bit and 411.2 GB at 4-bit: the weights (685.4B parameters) plus a working margin. A long context needs more. ### What is the cheapest GPU to run DeepSeek-V3.2 on? At 16-bit, 7x MI325X from $14.00 an hour; at 4-bit, 2x MI325X from $4.00 an hour, at the lowest on-demand prices the SAVRN Index lists. ### Can I use DeepSeek-V3.2 commercially? Yes. DeepSeek-V3.2 is released under MIT License. The MIT License is a short permissive license. It permits commercial use, modification and redistribution, provided the copyright notice and permission notice are included. ### What is DeepSeek-V3.2's context length? 163,840 tokens, from the maximum position embeddings in its published configuration. ## Similar Models Model · Text generation ### [DeepSeek-V3-0324](https://savrn.com/models/deepseek-v3-0324) [DeepSeek](https://savrn.com/model-publishers/deepseek-ai) DeepSeek-V3-0324 demonstrates notable improvements over its predecessor, DeepSeek-V3, in several key aspects. - More aesthetically pleasing web pages and game front-ends - Enhanced report analysis requests with more detailed outputs - Increased accuracy in Function Calling, fixing issues from previous V3 versions In the official DeepSeek web/app, we use the same system prompt with a specific date. For example, In our web and application environments, the temperature parameter $T{model}$ is set to 0.3. Because many users use the default temperature 1.0 in API call, we have implemented an API temperature $T{api}$ mapping mechanism that adjusts the input API temperature value of 1.0 to the… Open weights mit 684.5B parameters 163,840 tokens transformers [View model](https://savrn.com/models/deepseek-v3-0324) Model · Text generation ### [DeepSeek-R1](https://savrn.com/models/deepseek-r1) [DeepSeek](https://savrn.com/model-publishers/deepseek-ai) We introduce our first-generation reasoning models, DeepSeek-R1-Zero and DeepSeek-R1. DeepSeek-R1-Zero, a model trained via large-scale reinforcement learning (RL) without supervised fine-tuning (SFT) as a preliminary step, demonstrated remarkable performance on reasoning. With RL, DeepSeek-R1-Zero naturally emerged with numerous powerful and interesting reasoning behaviors. However, DeepSeek-R1-Zero encounters challenges such as endless repetition, poor readability, and language mixing. To address these issues and further enhance reasoning performance, we introduce DeepSeek-R1, which incorporates cold-start data before RL. DeepSeek-R1 achieves performance comparable to OpenAI-o1 across… Open weights mit 684.5B parameters 163,840 tokens transformers [View model](https://savrn.com/models/deepseek-r1) Model · Text generation ### [DeepSeek-V3](https://savrn.com/models/deepseek-v3) [DeepSeek](https://savrn.com/model-publishers/deepseek-ai) We present DeepSeek-V3, a strong Mixture-of-Experts (MoE) language model with 671B total parameters with 37B activated for each token. To achieve efficient inference and cost-effective training, DeepSeek-V3 adopts Multi-head Latent Attention (MLA) and DeepSeekMoE architectures, which were thoroughly validated in DeepSeek-V2. Furthermore, DeepSeek-V3 pioneers an auxiliary-loss-free strategy for load balancing and sets a multi-token prediction training objective for stronger performance. We pre-train DeepSeek-V3 on 14.8 trillion diverse and high-quality tokens, followed by Supervised Fine-Tuning and Reinforcement Learning stages to fully harness its capabilities. Comprehensive evaluations… Open weights 684.5B parameters 163,840 tokens transformers [View model](https://savrn.com/models/deepseek-v3) Model · Text generation ### [DeepSeek-V3-rescaled](https://savrn.com/models/deepseek-v3-rescaled) [Silares](https://savrn.com/model-publishers/sail-ing-north) We present DeepSeek-V3, a strong Mixture-of-Experts (MoE) language model with 671B total parameters with 37B activated for each token. To achieve efficient inference and cost-effective training, DeepSeek-V3 adopts Multi-head Latent Attention (MLA) and DeepSeekMoE architectures, which were thoroughly validated in DeepSeek-V2. Furthermore, DeepSeek-V3 pioneers an auxiliary-loss-free strategy for load balancing and sets a multi-token prediction training objective for stronger performance. We pre-train DeepSeek-V3 on 14.8 trillion diverse and high-quality tokens, followed by Supervised Fine-Tuning and Reinforcement Learning stages to fully harness its capabilities. Comprehensive evaluations… Open weights 684.5B parameters 163,840 tokens transformers [View model](https://savrn.com/models/deepseek-v3-rescaled) Model · Text generation ### [GLM-5.3](https://savrn.com/models/glm-5-3) [Z.ai](https://savrn.com/model-publishers/zai-org) GLM-5.3 uses the same base model as GLM-5.2 — every gain comes from post-training. Compared with GLM-5.2, it is much better at complex coding and long-horizon tasks: GLM-5.3 supports deployment with the following frameworks. Feel free to try them out: - SGLang — see cookbook - vLLM — see recipes - TokenSpeed — see here - Transformers — see transformers docs - KTransformers — see tutorial - Unsloth — see guide - For deployment on the Ascend NPU platform, inference frameworks such as vLLM-Ascend, xLLM and SGLang are supported — see here. - GLM-5.3 supports controlling the thinking budget through the reasoningeffort parameter, which accepts three levels: low, high, and max. It defaults to max… Open weights other 753.3B parameters 1,048,576 tokens transformers [View model](https://savrn.com/models/glm-5-3) Model · Text generation ### [GLM-5.2-FP8](https://savrn.com/models/glm-5-2-fp8) [Z.ai](https://savrn.com/model-publishers/zai-org) Join our WeChat or Discord community. Check out the GLM-5.2 blog and GLM-5 Technical report. Use GLM-5.2 API services on Z.ai API Platform. Try GLM-5.2 here. [ Paper ] [ GitHub ] We're introducing GLM-5.2, our latest flagship model for long-horizon tasks. It marks a substantial leap in long-horizon task capability over its predecessor GLM-5.1 and, for the first time, delivers that capability on a solid 1M-token context. GLM-5.2's new capabilities include: GLM-5.2 supports deployment with the following frameworks. Feel free to try them out: - SGLang (v0.5.13.post1+) — see cookbook - vLLM (v0.23.0+) — see recipes - Transformers (v0.5.12+) — see transformers docs - KTransformers (v0.5.12+)… Open weights mit 753.3B parameters 1,048,576 tokens transformers [View model](https://savrn.com/models/glm-5-2-fp8) ## DeepSeek [All models and datasets](https://savrn.com/model-publishers/deepseek-ai) ## Versions - [a7e62ac04ecb](https://savrn.com/models/deepseek-v3-2/versions/a7e62ac04ecb) · current 2026-10-23 ## Explore More - [All text generation models](https://savrn.com/models/tasks/text-generation) - [All models under mit](https://savrn.com/models/licenses/mit) - [Model comparisons](https://savrn.com/models/comparisons) - [The model directory](https://savrn.com/models) - [Open model prices by host](https://savrn.com/ai-index/pricing/open-models) ## Source - Repository metadata, read 2026-10-19. - [DeepSeek website](https://www.deepseek.com/) - [Hugging Face record](https://huggingface.co/deepseek-ai/DeepSeek-V3.2) - [How the hub is built](https://savrn.com/model-hub/methodology)