MMODELYST
Models/DeepSeek V3.2 Speciale
MDL

DeepSeek V3.2 Speciale

DeepSeek

Verified against Artificial Analysis · Jul 13, 2026
Capability
81
percentile index
Price /M
$0 / $0
in / out
Context
Avg score
58.8
10 benchmarks
AA Intelligence
22.2
index
Weights
Open
Released
Dec 1, 2025
About

DeepSeek V3.2 Speciale is a open-weights model in the DeepSeek family from DeepSeek. Benchmarked on 10 evals, averaging 58.8.

Benchmark scores · 10
avg 58.8 — higher is better
AIME
96.7
LiveCodeBench
89.6
GPQA Diamond
87.1
MMLU-Pro
86.3
IFBench
63.9
AA Long-Context Reasoning
59.3
SciCode
44
Terminal-Bench Hard
34.8
Humanity's Last Exam
26.1
τ²-bench
0
Details
PublisherDeepSeek
FamilyDeepSeek
Parameters
License
WeightsOpen
ReleasedDec 1, 2025
Capability over time

Append-only ledger — every observed change to this model's numbers.

API providers
Artificial Analysis$0/M