DeepSeek: DeepSeek V3.2 Exp
radar chart — all benchmarks
benchmark scores Scores from Mar 2026
| Categories | ||
|---|---|---|
| Coding | 33.3 | |
| Intelligence | 32.9 | |
| Agentic | 28.7 |
pricing — per 1M tokens via openrouter
Data unavailable
latency percentiles — time to first token (ms)
Data unavailable
model specifications
Context window —
Max output tokens —
Input modalities
—
Output modalities
—
Supports reasoning —
Supports tool use —