Gemini 3, released on November 18, 2025, is making waves for topping industry benchmarks, advanced reasoning and "vibe-coding" capabilities. But how does Google's latest model handle the specific challenges of web scraping?
On release day, we put Gemini 3.0 Pro to the test inside Zyte’s Web Scraping Copilot, our new Visual Studio Code extension designed to help data engineers build extractors faster.
The verdict: Top-tier quality
In our recent evaluation, Gemini 3.0 Pro produced the best results, delivering the most correct and concise web data extraction code.
It achieved the highest quality scores, improving on the already strong performance of Gemini 2.5 Pro and edging out leading models like GPT-5 and Claude.
The code it generates is very accurate and highly efficient. While other models often produce verbose or overly complex logic, Gemini 3 keeps it simple and effective.
The benchmarks: How we measured It
To see which model truly writes the best scraping code, we measured them across three key engineering metrics:
ROUGE-1 F1 (adjusted): ROUGE-1 F1 is a measure of the quality of generated output (0 to 1). It is our main metric for measuring code quality generated inside Web Scraping Pilot, adjusted here with smooth matching. Higher is better.
SLOC (Source Lines of Code): A measure of verbosity. We calculate how much executable code is generated per field. In scraping, concise code is generally more robust and easier to read and maintain. Lower is better.
Complexity: This measures the sophistication of the generated logic (number of decision paths per field). Simple, linear extraction logic is preferred over complex conditional spaghetti code. Lower is better.
gpt-5-mini
50.43
15.94
0.8027
gpt-5
38.71
13.68
0.8461
gpt-5.1
35.47
11.64
0.8414
gemini-2.5-pro
20.07
5.75
0.8469
haiku-4.5
19.11
5.62
0.7955
sonnet-4.5
20.66
6.00
0.7843
gpt-5.1-codex
35.61
12.10
0.8421
gemini-3-pro
21.49
6.28
0.8533
The catch: It’s still in preview
While the output quality is excellent, the experience isn't perfect yet. As Gemini 3.0 Pro is currently a preview model, we encountered stability issues during testing.
It sometimes produced empty output or timeouts, but not incorrect responses. We assume these issues will go away as the model enters general availability and Google scales it up according to demand.
In practice, we recommend waiting until it’s more mature.
Essentially, when it works, it works beautifully—but you might hit some bumps in the road until it scales up.
Scrape with Gemini 3.0 today
You can experiment with Gemini 3.0 Pro inside Web Scraping Copilot today to see its cutting-edge capabilities.
Web Scraping Copilot is a free Visual Studio Code extension for building and managing Scrapy spiders. It includes Zyte’s specialist scraping know-how that guides LLMs to generate the optimal scraping code that professionals need.
That includes auto-generating parsing code for target pages - a huge time-saver.
The best part? Web Scraping Copilot is model-agnostic. You aren't locked into a single LLM. If you prefer, you can use a reliable workhorse like GPT-5 or GPT-5-mini for your daily production workflows.
.gif&w=3840&q=70)
If you haven’t yet discovered Web Scraping Copilot, install it now:
Check out more Web Scraping Copilot content:
_HFpro5d6k3.png&w=256&q=75)
_E4PyVpfAxa.png&w=256&q=75)
-(1).png&w=1920&q=75)
-(1)_VZGHqxCgXV.png&w=1920&q=75)
