← DeepSeek V4 Pro · API · Max effort
Evidenceacceptedproduct_spec
Official model specification
DeepSeek API docs — models + first call
ev_2026_07_dsv4pro_spec
Result
1000000
Confidence S
- Tested
- 2026-04-24
Configuration
How it was measured
- Exact model
- deepseek-v4-pro
- Reasoning
- max
- Tool access
- openai_compatible_tools
- Prompting
- n/a
Caveats
Known limitations
- Text-only on this Build (no vision).
- Self-host quant path is a different Build if material.
Weighting
Evidence weight factors
The engine weights each evidence row by these factors (0–1) when fusing scores — higher is more trusted.
relevance1.00
quality0.70
recency0.95
health1.00
comparability1.00
reliability1.00
Benchmark
Official model specification
Official context window, modalities, reasoning modes, and API identifiers.
Benchmark source →