← DeepSeek V4 Pro · API · Max effort
Evidenceacceptedvendor_reported
SWE-bench Verified
DeepSeek V4 tech report / model card — SWE Verified
ev_2026_07_dsv4pro_swe_vendor
Result
80.6%
Confidence B
- Tested
- 2026-04-24
- Runs
- 1
Configuration
How it was measured
- Exact model
- deepseek-v4-pro
- Reasoning
- max
- Agent harness
- vendor_deepseek_swe
- Tool access
- agentic coding tools
- Prompting
- vendor published agent eval
Secondary STR — vendor SWE-bench Verified for V4-Pro-Max.
Caveats
Known limitations
- Vendor-reported; cannot receive confidence S.
- Harness ≠ mini-swe-agent openlm 76.2% row.
Weighting
Evidence weight factors
The engine weights each evidence row by these factors (0–1) when fusing scores — higher is more trusted.
relevance0.90
quality0.55
recency0.90
health0.80
comparability0.65
reliability0.70
Benchmark
SWE-bench Verified
Human-filtered 500-instance subset of SWE-bench evaluating whether models can resolve real GitHub issues in existing repositories under a controlled agent harness.
Benchmark source →