Skip to content

DeepSeek · Chat

DeepSeek V3.2 Speciale

DeepSeek 🇨🇳ChatDeepSeek familyOpen weights. One of its sources says the weights are publicly available.The maker currently lists 1 Nov 2027 as the deprecation date. Plans can change.
llminfo IQ
128.6
Market median price
$0.60 blended / 1M
Max context
164K
Listings
4

Evidence and sources

🤗 deepseek-ai/DeepSeek-V3.2-Specialellminfo IQ 128.6, ranked among chat models. From 33 benchmarks. Plenty of data behind it.How llminfo IQ is computedGood value (73.7). Cheaper per task than its benchmark results predict. From 10 benchmarks with measured task costs. Plenty of data behind it.How Value for money worksReleased 28 Nov 2025Maker HQ: CN (according to DeepSeek's own website)
Sources: LiteLLM data, FlowDot data, Models.dev data, Hugging FacePrimary data from LiteLLM from FlowDot: name from Models.dev: weights availability from Hugging Face: release date

Advantages and trade-offs

Compared with 10 chat models, picked mainly by popularity and IQ

Advantages

Open weights available

The weights are public. 2 of the 9 peers that say either way have open weights.

Trade-offs

We only make a claim when there's enough data on models of the same type. Updated 22 Nov, 06:07 UTC.

Input & output

as stated by the model's providers
Text
Text

Capabilities

Supported operations and provider-stated features

ReasoningFunction callingPrompt caching

Listings

 Microsoft Azureazure_ai/deepseek-v3.2-specialeSources: LiteLLM data, Models.dev data128K$0.58$1.68
 Aihubmixazure-deepseek-v3.2-specialeSource: AIHubMix data docsNot published by the provider$0.58$1.68
 Flowdotdeepseek/deepseek-v3.2-specialeSource: FlowDot data164K$0.40$1.20
 Gmicloudgmi/deepseek-ai/DeepSeek-V3.2-SpecialeSource: LiteLLM data164K$0.28$0.40

Benchmarks

Named effort levels run from lowest to highest. Select a result to see its source details.

Math

AIME 20252 results
Reported by DeepSeek · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
Speciale high-compute reasoning · temperature 1.0 · 128K context · pass@1
Source: LLM Stats data · Reported by the model's developer. The source relaying it did not verify it. claim
Setting not stated
HMMT3 related benchmarks · 3 results
HMMT 2025
Source: LLM Stats data · Reported by the model's developer. The source relaying it did not verify it. claim
Feb 25
Reported by DeepSeek · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
Speciale high-compute reasoning · temperature 1.0 · 128K context · pass@1
Nov 25
Reported by DeepSeek · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
Speciale high-compute reasoning · temperature 1.0 · 128K context · pass@1
IMOAnswerBenchSpeciale high-compute reasoning · temperature 1.0 · 128K context · pass@1 · Reported by DeepSeek · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.

Reasoning

Reported by DeepSeek · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
Speciale high-compute reasoning · temperature 1.0 · 128K context · pass@1
Source: LLM Stats data · Reported by the model's developer. The source relaying it did not verify it. claim
Setting not stated
t2-benchSource: LLM Stats data · Reported by the model's developer. The source relaying it did not verify it. claim

Coding

Codeforces RatingSpeciale high-compute reasoning · temperature 1.0 · 128K context · reported rating · Reported by DeepSeek · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
LiveCodeBenchSpeciale high-compute reasoning · temperature 1.0 · 128K context · pass@1-CoT · Reported by DeepSeek · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.

Agentic

General

Science

GPQA DiamondSpeciale high-compute reasoning · temperature 1.0 · 128K context · pass@1 · Reported by DeepSeek · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.

Other

CMO 2025Speciale high-compute reasoning · temperature 1.0 · 128K context · 102/126 · English version · generate-verify-refine · no tools/internet · Reported by DeepSeek · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
ICPC World Finals 2025Speciale high-compute reasoning · temperature 1.0 · 128K context · 10/12 solved · official time/attempt limits · no tools/internet · Reported by DeepSeek · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
IMO 2025 · Raw PointsSpeciale high-compute reasoning · temperature 1.0 · 128K context · 35/42 · generate-verify-refine · no tools/internet · official contest limits · Reported by DeepSeek · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
IOI 2025Speciale high-compute reasoning · temperature 1.0 · 128K context · 492/600 · official time/attempt limits · no tools/internet · Reported by DeepSeek · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.

Data credits: BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links. · Data by LLM Stats · EQ-Bench (Samuel J. Paech) — cite paech2023eqbench · MathArena (ETH SRI / INSAIT), MIT — cite dekoninck2026matharena (licence) · MLE-bench (OpenAI), MIT (licence) · Epoch AI Benchmarking Hub, CC BY 4.0 (licence) · SnakeBench (Greg Kamradt / ARC Prize Foundation), MIT (licence)

Also known as: azure-deepseek-v3.2-speciale, azure_ai/deepseek-v3.2-speciale, deepseek/deepseek-v3.2-speciale, gmi/deepseek-ai/deepseek-v3.2-speciale

Like this one. Tap or hover any dotted value, label or icon to see where it comes from and how it is measured. open details when tapped or hovered. How this data works