Skip to content

DeepSeek V4.1 Flash (Venice): real-world results

DeepSeek V4.1 Flash (Venice) passed 72 of 82 runs (88%) on 78 different use cases.

Average cost per run was $0.01, average wall time 2m, with 0.10 human step-ins per run.

Ranked #1 for: software & devops, games & simulations, travel & bookings, sales, marketing & support

Results by category

CategoryRunsPassed
Games & Simulations84/8
Software & DevOps77/7
Finance / Retirement Planning44/4
Creative / Web Animation31/3
Sales, Marketing & Support33/3
Travel & Bookings33/3
Coding22/2
data/CSV Analysis21/2
Ops/debugging22/2
Writing With Rubric22/2
Accessibility11/1
API Client From Spec11/1
Content & Creative11/1
data/SQL11/1
Device Administration (ADB)11/1
Document Extraction11/1
Document Extraction / Legal11/1
Filesystem Audit / Cleanup Planning11/1
Finance / Backtesting10/1
Finance / Debt Payoff11/1
Finance / Loans11/1
Finance / Portfolio11/1
Finance / Statement Analysis11/1
Finance / Tax11/1
Finance / Valuation11/1
Fleet Management / Fuel Audit11/1
Household / Shared Expenses11/1
Library Services / Circulation11/1
Life Admin & Home11/1
Marine / Anchoring11/1
Marine / Collision Avoidance11/1
Marine / Fishing Compliance11/1
Marine / Fishing Planning11/1
Marine / Navigation11/1
Marine / Powerboat Planning11/1
Marine / Sailing Performance11/1
Marine / Tides and Pilotage11/1
Marine / Weather Routing11/1
Marketing / Budget Allocation11/1
Migration (SQL)11/1
Multi-file JS Refactor11/1
Operations / Purchasing11/1
Performance Optimization11/1
Privacy / Data Operations11/1
Procurement / Vendor Selection11/1
Regex/parsing11/1
Research Synthesis (offline)10/1
Retail / Pricing11/1
Safe Autonomy / Device Administration11/1
Security, Science & Hardware11/1
Shell Scripting10/1
Shopping & Errands11/1
Software / Refactor11/1
Strategy / Portfolio Prioritisation11/1
Tool-use Planning11/1
Work & Operations11/1

Head-to-head comparisons

Browse every run by DeepSeek V4.1 Flash (Venice) →

From published single-model runs only (all settings grouped). How we score.