DeepSeek V3.1
ModelOpen sourceActiveby DeepSeek · family “deepseek-v”
DeepSeek-V3.1 is a large hybrid reasoning model (671B parameters, 37B active) that supports both thinking and non-thinking modes via prompt templates. It extends the DeepSeek-V3 base with a two-phase long-context... (description from the OpenRouter listing)
in: textout: textReasoningTool useStructured output
Epoch Capabilities Index
139.9 #113 of 268
90% CI 135.3 – 143.2
Released
21 Aug 2025
Context window
164K tokens
Max output
33K tokens
Input $ / 1M tokens
$0.25
Output $ / 1M tokens
$0.95
Cached input $ / 1M
$0.13
Each benchmark shown separately with its own source
Benchmark results (5)
| Benchmark | Domain | Score | vs best recorded | Setting | Run | Source |
|---|---|---|---|---|---|---|
| LMCA | agents | 24.3% | 37% | thinking | — | External ↗ |
| DTBench | reasoning | 82.7% | 84% | thinking | — | External ↗ |
| Fiction.LiveBench | long-context | 52.8% | 54% | — | — | External ↗ |
| WeirdML | coding | 38.4% ±0.0 | 41% | thinking | — | External ↗ |
| SimpleBench | reasoning | 40.0% | 49% | — | — | External ↗ |
Source: Epoch AI Benchmarking Hub (CC BY 4.0). “External” rows are leaderboard results Epoch collects from third parties. Best reported setting per benchmark is shown.
API list price over time
Price history
$ per 1M tokens
13 observations (OpenRouter listing + Internet Archive snapshots). History accumulates with every ingest run; a single point means no change has been observed yet.
Events
No events recorded yet.