DeepSeek-V3 vs Gemini 2.0 Flash

# benchmark_flash_vs_deepseek.py - Measure TTFT and token cost in real time
import time
import requests
from google import genai

# 1. Gemini 2.0 Flash via Google GenAI SDK
client = genai.Client(api_key="GEMINI_API_KEY")
t0 = time.time()
response = client.models.generate_content(
    model='gemini-2.0-flash',
    contents='Summarize PostgreSQL VACUUM strategies in 5 bullet points.'
)
gemini_duration = time.time() - t0
print(f"Gemini 2.0 Flash Latency: {gemini_duration:.3f}s | Output: {len(response.text.split())} words")

# 2. DeepSeek-V3 via OpenAI-compatible endpoint
t0 = time.time()
res = requests.post(
    "https://api.deepseek.com/chat/completions",
    headers={"Authorization": "Bearer DEEPSEEK_API_KEY"},
    json={
        "model": "deepseek-chat",
        "messages": [{"role": "user", "content": "Summarize PostgreSQL VACUUM strategies in 5 bullet points."}]
    }
).json()
deepseek_duration = time.time() - t0
print(f"DeepSeek-V3 Latency: {deepseek_duration:.3f}s | TTFT difference: {deepseek_duration - gemini_duration:.3f}s")