The story so far
Weights, every dated article, and repo commits on one rail.
17 repo commits since the weights landed
Teal is the weights (2025-07-30), grey ticks are repo commits.
dots.ocr
dots-studio MIT out 15 months ago
3.04B parameters, images and text in, text out, 128K tokens of context.
Verified 2026-10-25 Rank #438 in LLMs · score 78.3 method v1 Sources 3 Downloads 30d 505.4k · #165 of 619 tracked
Changed this week: 1 new quantized build on the Hub (6 in all).2026-10-24
Do people use it?505.4kdownloads, 30 days#165 of 619 we track · top 27%+2 likes, downloads −28% this week1,343 likesHugging Face2026-10-24
Where it's used
Hugging Face does not say who downloads a model or with what. These are the proxies: API providers listing it, quantized builds, fine-tunes/adapters/merges, and Spaces that load it.
No API provider serves it, so its downloads are people running it themselves: 6 quantized builds and 43 Spaces.
- As a service0no API provider listed
- On own hardware6quantized builds
- Built on10 fine-tunes · 1 adapters · 0 merges
- In apps43Hugging Face Spaces
Top Spaces running it: MohamedRashad/Dots-OCR ♥ 161 · prithivMLmods/Super-OCRs-Demo ♥ 44 · trevorpfiz/VLM-playground ♥ 5
Will it fit?
Weights only, an estimate: leave room for the KV cache. Method
Fits a 16 GB card at INT4
About 1.8 GB at INT4, 3.6 GB at FP8, 7.3 GB at BF16: weights only, so leave room for the KV cache.
Estimate params × bytes per param × 1.2 · how we estimate
What people built on it
7 derivatives on the Hub. Which build to run, by runtime: the most downloaded in each class.
7 derivatives on the Hub, most of them quantized
From the Hugging Face model tree; quantized builds in teal, fine-tunes, merges and adapters in greys.
Reference
Everything else, folded: the full tables, every source.
Benchmarks
12 rows
Benchmarks 12 rows
measured third-party leaderboard rowauthor-reported number from the model card or eval files (12)No confidence intervals unless the source gave them; settings differ between rows, so compare within a benchmark, not across. Date = the benchmark's own date when it gives one, else when we read it.
What the card says
2 cited sentences
Summary generated, cited
dots.ocr is a 3.04B parameter vision-language model designed for multilingual document layout parsing.[1][2] The model unifies layout detection and content recognition for tasks including text, tables, and formulas.[3]
- parameters human
- “Multilingual Document Layout Parsing in a Single Vision-L…”
- “unifies layout detection and content recognition within a…”
Generated 2026-10-07 15:09 UTC by parasail-gemma-4-26b-a4b-it; each numeral points at the fact it rests on. Not a verdict.
Use-case evidence
6 use cases
What it is good at 6 use cases
Labels are evidence, not verdicts: measured a third-party leaderboard row for this model; author-reported a number the model author published in the card or eval files; structural a deterministic signal from the name, config or chat template, or a decision-model estimate. Each chip links to its source. All use cases.
- Extract text from documentsauthor-reportedAuthor reports OmniDocBench 89.0decision-model estimate p=1.00name: ocr+1 more
- Understand images and videostructuraldecision-model estimate p=0.99tag: image-text-to-textpipeline: image-text-to-text+1 more
- Chat and follow instructionsstructuraltag: conversationalchat template present
- Run on a laptop or phonestructuraldecision-model estimate p=0.89fits: single-gpu
- Read long documentsstructuralconfig: 131,072 context
- Work across languagesstructuraltag: multilingual
Facts
21 fields
Listed in LLMs
Facts
- dots.ocr has 3.04B parameters. (safetensors metadata, observed 2026-10-24)
- Context length is 131,072 tokens (128K). (config.json max_position_embeddings, observed 2026-10-24)
- License: MIT (friction: none). Permissive. (license file, observed 2026-10-24)
- Created 2025-07-30, last modified 2025-10-31. (HF API, observed 2026-10-24)
| Pipeline tag | image-text-to-text |
|---|---|
| Library | dots_ocr |
| Architectures | DotsOCRForCausalLM |
| Model type | dots_ocr |
| Dtype | bfloat16 |
| Params by dtype | BF16: 3,039,179,264 |
| Size band | M |
| MoE | no |
| Context length | 131,072 (config.json max_position_embeddings) |
| Languages | en, zh, multilingual |
| License | MIT (mit) |
| Hub snapshot | 2026-10-25 |
Servability
vLLM · SGLang
Servability
vLLM: supported · DotsOCRForCausalLM · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: supported · DotsOCRForCausalLM · 773cb938fe9484a8f7e7e46fb1b1fe96412c59ed
vLLM registry (7050ef881ff0835f6890a47db1859a24c65481d6) · SGLang registry (773cb938fe9484a8f7e7e46fb1b1fe96412c59ed) · Matched by architecture class name against each project's registry on main; 'supported' means the architecture is registered, not that this checkpoint was tested.
Hardware fit
3 estimates
Hardware fit ESTIMATE
Weights-only memory for dots.ocr: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.
| BF16 (2 bytes/param) | 7.3 GB · Single GPU (<24 GB) |
|---|---|
| FP8 (1 byte/param) | 3.6 GB · Edge (<4 GB) |
| INT4 (0.5 bytes/param) | 1.8 GB · Edge (<4 GB) |
Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra.
All variants
7 variants
Lineage
quantized: 6adapter: 1
Top variants (7 shown, ranked by downloads)
| Variant | Relation | Quant | Params | Band | License | Author | Likes | Downloads |
|---|---|---|---|---|---|---|---|---|
| mradermacher/dots.ocr-i1-GGUF | quantized | GGUF | — | — | mit | mradermacher | 4 | 8.6k |
| helizac/dots.ocr-4bit | quantized | -4bit | 3.04B | M | mit | helizac | 31 | 5.4k |
| mradermacher/dots.ocr-GGUF | quantized | GGUF | — | — | mit | mradermacher | 5 | 4.9k |
| prithivMLmods/dots.ocr-GGUF | quantized | GGUF | — | — | mit | prithivMLmods | 2 | 1.9k |
| polymer/dots.mocr-GGUF | quantized | GGUF | — | — | apache-2.0 | polymer | 0 | 437 |
| samaritan-ai/dots-ocr-samaritan-hebrew | adapter | — | 3.04B | M | mit | samaritan-ai | 0 | 54 |
| hipfire-models/dots.ocr | quantized | — | — | — | mit | hipfire-models | 0 | 0 |
Repository changelog
17 commits
Repository changelog 17 commits
| Date | Commit | SHA |
|---|---|---|
| 2025-10-31 | Upload NOTICE with huggingface_hub | c0111ce |
| 2025-09-25 | Upload configuration_dots.py with huggingface_hub | ba670c5 |
| 2025-09-25 | Upload configuration_dots.py with huggingface_hub | b09554e |
| 2025-09-25 | fix autoprocessor (#30) | 6b68a58 |
| 2025-09-22 | Adding `transformers` as a library, and also mentioning the `custom_code` tag (#29) | 89eabff |
| 2025-09-13 | Update VisionSdpaAttention to support memory efficient backend. (#27) | 4d84cbc |
| 2025-08-18 | Update modeling_dots_vision.py (#19) | 325ed02 |
| 2025-08-12 | Upload dots.ocr LICENSE AGREEMENT with huggingface_hub | e315fc3 |
| 2025-08-07 | Upload NOTICE with huggingface_hub | bb6e753 |
| 2025-08-07 | Update modeling_dots_ocr_vllm.py (#7) | bb9eb0e |
| 2025-08-04 | Upload modeling_dots_vision.py with huggingface_hub | 42f4178 |
| 2025-07-31 | Upload README.md with huggingface_hub | f984833 |
| 2025-07-31 | Upload README.md with huggingface_hub | 3d2aea7 |
| 2025-07-30 | Upload README.md with huggingface_hub | ebfa26a |
| 2025-07-30 | Upload README.md with huggingface_hub | aaccbca |
| 2025-07-30 | Upload folder using huggingface_hub | e1cb4af |
| 2025-07-30 | initial commit | 7638595 |
Similar models
3 models
Similar models
- paligemma-3b-pt-224 · 2.92B same size band (M), same task (image-text-to-text)
- paligemma2-3b-pt-224 · 3.03B same size band (M), same task (image-text-to-text)
- d1-3B · 3.12B same size band (M), same task (image-text-to-text)
Sources
3 sources
Sources
Hub snapshot: 2026-10-25 · HF API: 2026-10-24 21:38 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:46 UTC · Summary: 2026-10-07 15:09 UTC · Last verified 2026-10-25
Follow dots.ocr
Watch this model
- /feeds/models/dots-studio--dots.ocr.xml(RSS 2.0) · every change to dots.ocr
- /feeds/families/dots-studio--dots.ocr.xml(RSS 2.0) · every change across the dots.ocr family
Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.