DeepSeek V4 Flash Vision (Reasoning, Max Effort)

DeepSeekopentext

217.1Speed (tok/s) · P93 · measured
$0.227Blended $/M tok · P69 · derived
35Intelligence index · P86 · measured
107 msTTFT · measured
—Context
65Coding index · measured
48Agentic index · measured

Embed this card (static and self-contained; updates when the site deploys — public deploys are approval-gated):

<iframe src="https://viz.kyanitelabs.tech/embed/deepseek-v4-flash-vision-reasoning-max-effort" width="100%" height="150" style="border:1px solid #e3e8ef;border-radius:10px" title="DeepSeek V4 Flash Vision (Reasoning, Max Effort) — Kyanite Labs model card"></iframe>

Comparisons (measured vs nearest neighbors): GPT-5.6 Luna (xhigh) · Grok 4.6 (low)

Released 2026-08-21 · catalog data 2026-09-14 · field-level provenance shown per metric above (measured / derived / list, from the catalog sources map); percentiles computed across the 329-model catalog at generation time. Numbers are catalog-sourced and labeled as such — lab-measured lanes stay separate by design. Generated page — do not hand-edit.

Model names are trademarks of their respective owners. Not affiliated with, endorsed by, or sponsored by any provider or measurement service. Sources, attribution & corrections.