DeepSeek V4 Pro 0813
Pinned 0813 DeepSeek V4 Pro checkpoint for deterministic long-running agent deployments.
Dated 0813 snapshot of DeepSeek V4 Pro delivering repeatable 1T sparse-MoE reasoning behavior across long-context coding and analysis pipelines without snapshot drift.
Designed for production agent harnesses and evaluation suites that require a pinned checkpoint. Shares V4 Pro 1M context and vision capabilities with predictable tool-calling format across runs.
Pinned production agentsReproducible evalsLong-context reasoningPinned agent pipelinesRegression suitesDeterministic code review
Strengths
Reproducible prompt and tool behavior, full 1M context window, high coding and formal reasoning accuracy.
Considerations
Use `deepseek-v4-pro-0813` when pinning CI/CD or autonomous agent behavior to avoid unannounced model changes.
Live throughput history
source: status.ebotservers.comParameters1T
Context1M
Max output384K
Intelligence index23
TPS avg 24h0.0
TTFB p500 ms
Uptime100.0%
Call it
Other models from DeepSeek
A new version of this page is available