DeepSeek 4.1 Flash speed on the official API: about 325 tokens/s

Melvin Vivas · X post · 2026-09-14 · Open on X

Topics: LLM Fundamentals, LLMOps, Deployment & Monitoring · Level: beginner

Summary

The creator reports that DeepSeek 4.1 Flash generates about 325 tokens per second through DeepSeek's official API. This is one data point to keep in mind when comparing models on latency and speed.

Key points

• DeepSeek 4.1 Flash ran at about 325 tok/s on DeepSeek's official API
• This is a single observed measurement from one user, not a formal benchmark
• Output speed (tokens per second) is a useful thing to compare when picking a model for latency-sensitive apps

Resources mentioned

• DeepSeek V4.1 Flash · tool · api-docs.deepseek.com · paid
  A fast DeepSeek language model available through DeepSeek's official API.
  Also in: Using Hugging Face credits: Jobs, Inference Endpoints and Open Models (Melvin Vivas on X · notes), One-Shot Agent Management UI with DeepSeek Harness for $0.19 (Melvin Vivas on X · notes), DeepSeek Harness: Open-Source, Browser-Based Agent for DeepSeek V4.1 (Melvin Vivas on X · notes), DeepSeek V4.1 Flash Off-Peak Pricing as a Cheap Fallback Model (Melvin Vivas on X · notes) and 1 more
• DeepSeek API · tool · platform.deepseek.com · paid
  DeepSeek's official API platform for buying credits and managing API keys.
  Also in: Getting Started with DeepSeek Harness and the DeepSeek API Platform (Melvin Vivas on X · notes), DeepSeek-V4-Flash Official API Launches in Public Beta (Melvin Vivas on X · notes)