Best Coding Model
DeepSeekProvisional

DeepSeek-V3.2-Speciale

DeepSeek / DeepSeek-V3.2-Speciale · deepseek-ai/DeepSeek-V3.2-Speciale

High-compute DeepSeek reasoning variant for evaluation and research-heavy tasks.

Objective score

56.1

Confidence

85%

Price evidence

Pricing evidence pending

Support this modelCompare trendOfficial source

Other primary tiers in this family

DeepSeek R1 0528 Qwen3 8BDeepSeek R1 Distill Llama 70BDeepSeek R1 Distill Llama 8BDeepSeek R1 Distill Qwen 1.5BDeepSeek R1 Distill Qwen 14BDeepSeek R1 Distill Qwen 32BDeepSeek R1 Distill Qwen 7BDeepSeek V2.5 1210DeepSeek-R1-0528DeepSeek-V3-0324DeepSeek-V3.2DeepSeek-V3.2-Speciale

Evidence state

Signal coverage

Benchmarks
2
Rank points
7
Usage
0
Agent
0
Sources
3

Objective evidence

Benchmarks and public signals

Swebench
60
Livecodebench
—
Aider
74.2
Artificial Analysis
—
Openrouter Usage
—
Kilo Usage
—

Persistent history

Recent ranking points

Aug 1#4 · 67
Aug 1#4 · 67
Jul 31#4 · 67
Jul 31#4 · 67
Jul 31#4 · 67
Jul 31#4 · 67

Usage signals

Evidence is being collected. Missing data is not treated as zero.

Agent and tool signals

Evidence is being collected. Missing data is not treated as zero.