need a model for ocr or vqa? try gemini 3.5 flash. gemini 3.5 flash is faster, cheaper, and more accurate.
Details ↓
5:23 AM · Jul 6, 2026 · 17.1K Views
Sentiment
Positive users highlight Gemini 3.5 Flash's speed, cost and accuracy on VQA and OCR tasks while negative users call it inferior to prior versions or not cheap enough compared with alternatives.
Pos
42.9%
Neg
57.1%
7 comments with sentiment.
Cluster Engagement
Digg Deeper
No Digg Deeper questions have been answered for this story yet.
Posts from X
Most Activity
Most Activity
VIEWS2.4KBOOKMARKS4LIKES7RETWEETS2
Philipp Schmid@_philschmid
SkalskiP@skalskip92
need a model for OCR and VQA? Fable 5 is NOT your best option
Gemini 3.5 Flash is 1.3x faster, 2.9x cheaper, and more accurate
Fable 5 struggles especially on tasks that require counting objects.
↓ cost comparison and interesting failure cases
4hViews 2.4KLikes 7Bookmarks 4
REPLIES1
avi@aviaip
@_philschmid why not 3.1 flash lite? OCR is a lighweight task isn't it?
2hViews 232Likes 1
Howard@0xHoward_Peng
@_philschmid why no deepseek OCR?
3hViews 346Likes 1Bookmarks 1
The AI Bench@theaibenchai
@_philschmid Worth checking how it handles dense tables and rotated scans specifically, that's where most OCR benchmarks fall apart even when aggregate accuracy looks great. Anyone run it against real invoice/receipt datasets yet, not just doc-VQA leaderboards?
3hViews 74Likes 2
dan.@danrodrigux
@_philschmid Absolutely real
2hViews 183
Harris7@KrishWiller
@_philschmid 一点也不便宜
2hViews 153
AcroTic@AcroTicRL
@_philschmid Honestly, in my experience Gemini 3 Pro had the best vision and understanding I'd ever seen. It's a shame you discontinued it. 3.5 Flash is still struggling a lot more than 3.1 Pro. Idk what the benchmarks say
1hViews 147
Icarus@Icarussiano
@aviaip @_philschmid But vqa isnt
2hViews 29Likes 1
ANIRUDDHA ADAK@aniruddhadak
@_philschmid ok, but when 3.5 pro ?
1hViews 44
Viswa Reddy@lviswanath
@_philschmid Yes, I tried Gemini 3.5 flash for OCR. It is fast, cheaper and accurate than others.
I built an AI agent and workflow for using Google drive, Gemini flash and Gmail for a Accounting client, CA and a Business owner.
It works like a charm.
1hViews 42
林夕AI@xilin4096986838
@_philschmid paddleocr better believe me
56mViews 10
worlder@T1443705763507
@_philschmid 3.5 pro发布之前，可以闭嘴哈哈
4hViews 6
veloX@veloXxn
@_philschmid gemini 3.5 flash just became the default ocr/vqa choice and everyone's switching from claude fable within a week. speed tier matters more than people thought.
1hViews 2