VLM Run (@vlmrun) header
VLM Run (@vlmrun) profile picture

VLM Run

@vlmrun

Visual Intelligence for Developers

 Santa Clara, CA
vlm.run/
Joined January 2022
  • Tweets252
  • Following52
  • Followers696
68Media
Photo by VLM Run (@vlmrun)Photo by VLM Run (@vlmrun)Photo by VLM Run (@vlmrun)Photo by VLM Run (@vlmrun)Photo by VLM Run (@vlmrun)Photo by VLM Run (@vlmrun)Photo by VLM Run (@vlmrun)Photo by VLM Run (@vlmrun)Photo by VLM Run (@vlmrun)Photo by VLM Run (@vlmrun)
Pinned Tweet
VLM Run (@vlmrun) profile picture
VLM Run
@vlmrunSep 1
Introducing VLM Run Gateway: run open-weight VLMs, OCR VLMs, embodied VLMs, and specialized ViTs in one OpenAI-compatible API. You don't need a trillion parameters to parse documents, understand images, or summarize videos. When every model runs on one API, changing models is as simple as changing a parameter. The visual frontier just moved. Your AI bill should too. $ uvx vlmrun gw models
9
6
4
39
646,926
New: EmbeddingGemma 2 on the @vlmrun Gateway. One model for text, image, document, video, and audio embeddings. 740M parameters, 768 to 128 dimensions, $0.01 per 1M tokens.
EmbeddingGemma 2, now on VLM Run Gateway. Text, images, documents, video, and audio in one vector space.

ALT EmbeddingGemma 2, now on VLM Run Gateway. Text, images, documents, video, and audio in one vector space.

One embedding space: a map of twelve inputs, each a dot. An invoice image and an invoice PDF sit together with the text an invoice; two different robot videos sit together with the text robot arms; a video about a bakery, a photo of a vintage car and an audio clip each sit next to their text. Images, video, audio, PDFs and text from a live EmbeddingGemma 2 run.

ALT One embedding space: a map of twelve inputs, each a dot. An invoice image and an invoice PDF sit together with the text an invoice; two different robot videos sit together with the text robot arms; a video about a bakery, a photo of a vintage car and an audio clip each sit next to their text. Images, video, audio, PDFs and text from a live EmbeddingGemma 2 run.

Model card: one multimodal embedding model. Text, code, image, document, video, and audio, searched together in one vector space. $0.01 per 1M tokens, 740M total params, 768 to 128 dims picked per request, Apache 2.0 license.

ALT Model card: one multimodal embedding model. Text, code, image, document, video, and audio, searched together in one vector space. $0.01 per 1M tokens, 740M total params, 768 to 128 dims picked per request, Apache 2.0 license.

Get started in a few lines of code: Python calling client.embeddings.create with model google/embeddinggemma-2, an image input, and prompt_name Document, printing 768 dims and 288 tokens; commented lines show the same call with an audio file (speech.wav), a PDF (report.pdf) or a video (clip.mp4). From your terminal: uvx vlmrun gw embed for an image, a video clip, and a text query. Sign up for a free API key at app.vlm.run.

ALT Get started in a few lines of code: Python calling client.embeddings.create with model google/embeddinggemma-2, an image input, and prompt_name Document, printing 768 dims and 288 tokens; commented lines show the same call with an audio file (speech.wav), a PDF (report.pdf) or a video (clip.mp4). From your terminal: uvx vlmrun gw embed for an image, a video clip, and a text query. Sign up for a free API key at app.vlm.run.

1
2
5
193
VLM Run model card: vlm.run/gateway/models/googl… Hugging Face model card: huggingface.co/google/embedd… EmbeddingGemma 2 developer guide: developers.googleblog.com/em…
EmbeddingGemma 2 | VLM Run

EmbeddingGemma 2 | VLM Run

Multimodal embeddings: text, images, video, audio, and PDF pages in one 768-d space.

vlm.run
1
2
47
New: Our System One API can now extract JSON fields, fast! Text, numbers, dates, and lists, each read in parallel and returned typed, or null when it's missing. Up to 3.7x faster than constrained JSON schema extraction, and up to 3.3x faster than JSON mode, on the same model. Support for native PDFs (up to 16 pages), images, and text inputs.
Bar chart of p50 latency to extract 10 invoice fields with Gemma 4 26B, /chat/completions vs /systemone. Text: System One 308 ms, JSON Schema 1,132 ms, JSON Mode 1,012 ms. Two-page PDF: 410 ms, 1,260 ms and 1,104 ms.

ALT Bar chart of p50 latency to extract 10 invoice fields with Gemma 4 26B, /chat/completions vs /systemone. Text: System One 308 ms, JSON Schema 1,132 ms, JSON Mode 1,012 ms. Two-page PDF: 410 ms, 1,260 ms and 1,104 ms.

2
2
1
5
478
API Docs: docs.vlm.run/gateway/system-… Extraction Guide: docs.vlm.run/gateway/guides/…
System One - VLM Run

System One - VLM Run

Typed, calibrated decisions over text, JSON, and images in a single read

docs.vlm.run
1
40
New: LightOnOCR-3-4B on the @vlmrun Gateway. One model for Markdown, grounding, and layout. 4B params, 86.3 on olmOCR-Bench, 75.1 on ParseBench.
LightOnOCR-3-4B, now on VLM Run Gateway. One model for Markdown, grounding, and layout.

ALT LightOnOCR-3-4B, now on VLM Run Gateway. One model for Markdown, grounding, and layout.

Every value comes with its box: a sample invoice with the model's grounding output drawn on it, each block boxed and labeled as title, text, table or footer.

ALT Every value comes with its box: a sample invoice with the model's grounding output drawn on it, each block boxed and labeled as title, text, table or footer.

Model card: three methods, markdown, grounding and layout, with block types text, title, table, chart, formula, image, footer. $0.20 per 1M input and $0.50 per 1M output tokens. 4B params, Apache 2.0. 86.3 on olmOCR-Bench, 75.1 on ParseBench.

ALT Model card: three methods, markdown, grounding and layout, with block types text, title, table, chart, formula, image, footer. $0.20 per 1M input and $0.50 per 1M output tokens. 4B params, Apache 2.0. 86.3 on olmOCR-Bench, 75.1 on ParseBench.

Get started in a few lines of code: Python calling client.chat.completions.create with model lightonai/lightonocr-3-4b and method grounding, printing the blocks as JSON. From your terminal: uvx vlmrun gw chat page.jpg with methods markdown, grounding and layout. Sign up for a free API key at app.vlm.run.

ALT Get started in a few lines of code: Python calling client.chat.completions.create with model lightonai/lightonocr-3-4b and method grounding, printing the blocks as JSON. From your terminal: uvx vlmrun gw chat page.jpg with methods markdown, grounding and layout. Sign up for a free API key at app.vlm.run.

1
1
3
153
VLM Run model card: vlm.run/gateway/models/light… Hugging Face model card: huggingface.co/lightonai/Lig… LightOnOCR-3 blog: huggingface.co/blog/lightona…
LightOnOCR-3 4B | VLM Run

LightOnOCR-3 4B | VLM Run

Page to Markdown, or labeled blocks with boxes in one call.

vlm.run
1
50
I made a computer vision demo for push ups! Working with @movewithmigs on strength training, and I did a push ups test to see how many I could do. I thought I did 25, but computer vision counted 24 🤣 keeping me honest lol I'm using ViTPose+ Large through the @vlmrun Gateway for pose estimation. I'm using the vertical displacement of the shoulder to detect full range of motion, and using the ankle-hip-shoulder line to compare against a straight back reference. You can see a plot of how much my back deviates from the straight line reference. In the last two reps, you can see that I am deviating a bit. Since everyone’s push ups look different, we can tune the thresholds accordingly if needed. Let me know what you think! The code is open-source on GitHub: github.com/jeremyipark/visio…
21
11
1
212
6,734
VLM Run retweeted
VLM Run (@vlmrun) profile picture
VLM Run
@vlmrunOct 14
Run and compare open-weight OCR models on the new @vlmrun OCR Playground. Auto-eval and compare costs between several VLMs including PP-OCRv6, GLM-OCR, PaddleOCR-VL 1.6, DeepSeek-OCR-2, and the new LightOnOCR-3! Try the OCR Playground on your own data: chat.vlm.run/#ocr
Image from VLM Run (@vlmrun)
2
5
30
2,673,813
Run and compare OCR models side by side on the @vlmrun OCR Playground. For models with grounding, hover any extracted value to see exactly where it came from on the page. Models Supported: PP-OCRv6, LightOnOCR-3-4B, dots.mocr, PaddleOCR-VL-1.6, GLM-OCR, DeepSeek-OCR-2 and more.
1
4
1
7
344
Try the OCR Playground on your own data: chat.vlm.run/ocr
OCR Playground - VLM Run

OCR Playground - VLM Run

Compare OCR and document AI models side-by-side on PDFs and images.

chat.vlm.run
1
65
We just added @typesafeai's native Image and Box support to our System One API. Pass an Image as the state, ask Box questions, get the region back as four edges. Same TypeSafe SDK: point it at gateway.vlm.run/typesafe.
Image from VLM Run (@vlmrun)
Image from VLM Run (@vlmrun)
1
2
1
6
474
Docs: docs.vlm.run/gateway/system-…
System One - VLM Run

System One - VLM Run

Typed, calibrated decisions over text, JSON, and images in a single read

docs.vlm.run
1
58
Run and compare open-weight OCR models on the new @vlmrun OCR Playground. Auto-eval and compare costs between several VLMs including PP-OCRv6, GLM-OCR, PaddleOCR-VL 1.6, DeepSeek-OCR-2, and the new LightOnOCR-3! Try the OCR Playground on your own data: chat.vlm.run/#ocr
Image from VLM Run (@vlmrun)
2
5
30
2,673,813
Yes, @LightOnIO models are really good! So is dots.mocr.
590
VLM Run retweeted
Sudeep Pillai (@spillai) profile picture
Sudeep Pillai
@spillaiOct 11
We're heading into a new era of sports analytics, where any cricket enthusiast or aspiring athlete can analyze their own personal videos and start building an AI-powered personal coach. I spent about 10 minutes fetching clips, extracting bowling form with the @vlmrun gateway and comparing bowling traces from @Jaspritbumrah93, @malinga_ninety9 and others.
2
6
2
62
3,861
New: System One reads pixels. Ask a typed question about an image. Get a calibrated distribution over your labels back, not a string to parse. Works with the TypeSafe SDK unchanged: point base_url at gateway.vlm.run/typesafe
22 numbered lines of Python. A TypeSafeClient for gateway.vlm.run/typesafe with model google/diffusiongemma-26b-a4b-it reads invoice.png as a data URL and asks System One "What kind of page is this?" with choices invoice, receipt and slide, then prints the label.

ALT 22 numbered lines of Python. A TypeSafeClient for gateway.vlm.run/typesafe with model google/diffusiongemma-26b-a4b-it reads invoice.png as a data URL and asks System One "What kind of page is this?" with choices invoice, receipt and slide, then prints the label.

2
1
6
304
Docs: docs.vlm.run/gateway/system-…
System One - VLM Run

System One - VLM Run

Typed, calibrated decisions over text, JSON, and images in a single read

docs.vlm.run
2
87
September changelog. Lots of exciting updates last month, and lots more to come including new guides and cookbooks.
VLM Run September 2026 changelog: gateway beta with 20+ visual models; new System One (s1) API, TypeSafe-compatible, image inputs; TypeSafe SDK works by swapping base_url; s1 engines DiffusionGemma-26B, Gemma4-26B, Qwen3.8-27B; WebSocket for s1 decisions; new gateway docs.

ALT VLM Run September 2026 changelog: gateway beta with 20+ visual models; new System One (s1) API, TypeSafe-compatible, image inputs; TypeSafe SDK works by swapping base_url; s1 engines DiffusionGemma-26B, Gemma4-26B, Qwen3.8-27B; WebSocket for s1 decisions; new gateway docs.

1
1
5
219
Docs: docs.vlm.run/gateway/changel…
Changelog - VLM Run

Changelog - VLM Run

Monthly Gateway highlights: models, features, billing, and limits.

docs.vlm.run
2
103
I made a computer vision tool for rock climbing holds analysis! As someone who is newer to climbing and wants to improve, I've noticed more and more recently how everyone solves the route differently. As I said to my friends the other day, "Everyone has to find their own way." 🤣 So I extended my last rock climbing demo with new analytics that I'm really excited about. I added what percentage of the total time your hands and feet are on a hold, as well as which holds each hand and foot is using. This data can be used in a few ways. For example, I know that I should use my feet and legs more when climbing. This tool can show me how much I use my feet, so I can use that as a baseline to improve. What I'm most excited about is how it enables comparing climbing sequences on the same route. You can compare against your friends' sequences: maybe you use 1, 3, 5, 7, and another friend uses 1, 2, 6, 7. Or you can compare against your own historical data, like I show here at the end. I used ViTPose+ Large for pose estimation and SAM 3.1 to segment the holds, both models accessed through the @vlmrun Gateway. I also use SAM 3.1 to segment the gray floor, so we know when the feet leave the ground to help with starting the route timer. Would love to hear what you think! The code is open-source on GitHub: github.com/jeremyipark/visio…
42
80
10
782
72,083
Load more