Roboflow (@roboflow) / X

Roboflow

2,126 posts

Roboflow

@roboflow

Build & use computer vision models fast ✨ Get started: roboflow.com Open source datasets & models universe.roboflow.com

Joined June 2019

Roboflow
@roboflow
Jul 30
pretty clear at this point what model you should use to automate labeling vision data
SkalskiP
@skalskip92
Jul 30
“why use VLMs for open-vocabulary detection? YOLO-E also supports text prompts and runs in real time.” true. but text prompting is not the same as language understanding using the same class names as for VLMs, YOLOE-26x scored 20.2% mAP@50 and YOLOE-11l scored 18.3% prompts
Roboflow
@roboflow
Jul 29
just expanded our image reasoning evals and gemini 3.5 flash is still at the top best frontier model for visual tasks without a doubt we've benchmarked all the latest models and google takes the top 3 spots (and they are the fastest!) followed by muse spark 1.1 then fable 5
SkalskiP
@skalskip92
Jul 29
gemini 3.5 flash is sooooooo good at image reasoning image reasoning is when the model has to actually think about what it sees not just label objects or read text it needs to connect the facts and figure out what’s going on reasoning trace: current order top to bottom: 654,
Roboflow
@roboflow
Jul 22
gemini 3.6 flash breakdown with example images key finding: 3.5 Flash-Lite beats Fable 5 while being ~25x cheaper gemini models continue to be the best models for vision use cases, way cheaper and faster than any other model
SkalskiP
@skalskip92
Jul 22
I’m both impressed and disappointed by Gemini 3.6 Flash faster, cheaper, and uses fewer tokens then Gemini 3.5 Flash. noticeably worse at object detection. often returns one general box instead of several precise ones. feels lazy. prompt: detect banana tree ↓ deep dive
Roboflow
@roboflow
Jul 21
we benchmarked inkling on vision tasks, it's coming in at the bottom of the leaderboard. probably expected given they are leaning into fine-tuning, nothing interesting there what is interesting: it is so clear common benchmarks are saturated just a few percentage different
Thinking Machines
@thinkymachines
Jul 15
Today, we are introducing Inkling. Inkling reasons efficiently across text, image, and audio modalities. We are making the full weights available. thinkingmachines.ai/news/introduci… Available today for fine-tuning on Tinker. Play with it in the Inkling Playground. 🧵
Roboflow
@roboflow
Jul 2
very good thread with sample data showing Fable 5 performance against other models for vision tasks important to see the types of tasks frontier models can/cant do will give you ideas for how to use them successfully in production
SkalskiP
@skalskip92
Jul 2
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
See @roboflow's full profile
Sign up
Log in

See @roboflow's full profile

See @roboflow's full profile