Unspecified · LIVEBENCH 2026-06-25

deepseek-v4-flash

Published benchmark results for this specific model configuration.

deepseek-v4-flash

Overall score65.5 / 100

Cost / successful task$0.016USD · benchmark workload

Weight accessNot reported

Where this configuration performs

Category averages from the same release, on a 0–100 scale.

Reasoning70.6

Coding69.2

Agentic Coding37.6

Mathematics79.6

Data Analysis68.0

Language70.1

Instruction Following63.1

Task-level results

theory of mind73.1zebra puzzle49.3spatial96.0logic with navigation64.0code generation73.2code completion65.2javascript54.5typescript23.3python35.0AMPS Hard98.0integrals with game37.0math comp97.1olympiad86.5consecutive events59.9tablejoin46.1tablereformat98.0connections89.5plot unscrambling46.9typos74.0paraphrase62.4simplify54.0story generation67.3summarize68.8

Source files: Scores ↗ Categories ↗ Costs ↗ . Missing cost is unknown, never free. Compare configurations within the same release.