Model openness
What each model makes available.
Six icons describe each model’s weights, license, training code, training data, paper and documentation, following Danish ASR and MTEB. Red icons mark confirmed availability. Hover over an icon on the leaderboard for an explanation, or select it to read the evidence.
What each icon means
- Weights
- Downloadable model weights, including releases that require accepting terms.
- Open license
- A license permitting use, modification and redistribution, including commercial use, without field-of-use restrictions.
- Training code
- Public code and a recipe for the training that produced this checkpoint; inference code and generic fine-tuning tutorials are insufficient.
- Training data
- The complete training corpus is publicly available, or can be reconstructed from public sources and a released recipe. A description alone is insufficient.
- Paper / report
- A public technical paper or report describing this model.
- Model documentation
- Public documentation of the model, its intended use and how to run it.
Checks apply to the evaluated model’s published materials. Faded icons can mean partial availability, a closed release, or insufficient information; each is labelled separately in the evidence. The criteria are not combined into a score.
These are ScandiOCR’s assessments. Openness does not affect OCR scores or rankings.
Riksarkivet/trocr-base-handwritten-hist-swe-2
| Check | Status | Evidence |
|---|---|---|
| Weights | Available | Downloadable checkpoint files. |
| Open license | Available | Apache-2.0 model license. |
| Training code | Unknown | The code and complete training recipe for this model have not been verified. |
| Training data | Partial | The card links public training datasets but explicitly states that not all training data can be released. |
| Paper / report | Unknown | A technical paper about this model has not been verified. |
| Model documentation | Available | Public model card. |
Reviewed 2026-10-13 · Link to this model
Sprakbanken/TrOCR-norhand-v3
| Check | Status | Evidence |
|---|---|---|
| Weights | Available | Downloadable checkpoint files. |
| Open license | Available | CC-BY-4.0 model license. |
| Training code | Unknown | The code and complete training recipe for this model have not been verified. |
| Training data | Available | The card identifies NorHand v3 as the training dataset for this release and links its public version. |
| Paper / report | Unknown | A technical paper about this model has not been verified. |
| Model documentation | Available | Public model card. |
Reviewed 2026-10-13 · Link to this model
stepfun-ai/GOT-OCR-2.0-hf
| Check | Status | Evidence |
|---|---|---|
| Weights | Available | Downloadable checkpoint files. |
| Open license | Available | Apache-2.0 model license. |
| Training code | Partial | The authors release training code using GOT weights. This does not establish a complete recipe for the evaluated checkpoint. |
| Training data | Unknown | A complete public training corpus or reconstruction recipe has not been verified. |
| Paper / report | Available | A technical report describes this model family. |
| Model documentation | Available | Public model card. |
Reviewed 2026-10-13 · Link to this model
zai-org/GLM-OCR
| Check | Status | Evidence |
|---|---|---|
| Weights | Available | Downloadable checkpoint files. |
| Open license | Available | MIT model license. |
| Training code | Unknown | The code and complete training recipe for this model have not been verified. |
| Training data | Unknown | A complete public training corpus or reconstruction recipe has not been verified. |
| Paper / report | Available | A technical report describes this model family. |
| Model documentation | Available | Public model card. |
Reviewed 2026-10-13 · Link to this model
PaddlePaddle/PaddleOCR-VL-1.6
| Check | Status | Evidence |
|---|---|---|
| Weights | Available | Downloadable checkpoint files. |
| Open license | Available | Apache-2.0 model license. |
| Training code | Unknown | The code and complete training recipe for this model have not been verified. |
| Training data | Unknown | A complete public training corpus or reconstruction recipe has not been verified. |
| Paper / report | Available | A technical report describes this model family. |
| Model documentation | Available | Public model card. |
Reviewed 2026-10-13 · Link to this model
Qwen/Qwen2-VL-2B-Instruct
| Check | Status | Evidence |
|---|---|---|
| Weights | Available | Downloadable checkpoint files. |
| Open license | Available | Apache-2.0 model license. |
| Training code | Unknown | The code and complete training recipe for this model have not been verified. |
| Training data | Unknown | A complete public training corpus or reconstruction recipe has not been verified. |
| Paper / report | Available | A technical report describes this model family. |
| Model documentation | Available | Public model card. |
Reviewed 2026-10-13 · Link to this model
Qwen/Qwen3-VL-2B-Instruct
| Check | Status | Evidence |
|---|---|---|
| Weights | Available | Downloadable checkpoint files. |
| Open license | Available | Apache-2.0 model license. |
| Training code | Unknown | The code and complete training recipe for this model have not been verified. |
| Training data | Unknown | A complete public training corpus or reconstruction recipe has not been verified. |
| Paper / report | Available | A technical report describes this model family. |
| Model documentation | Available | Public model card. |
Reviewed 2026-10-13 · Link to this model
Qwen/Qwen3-VL-4B-Instruct
| Check | Status | Evidence |
|---|---|---|
| Weights | Available | Downloadable checkpoint files. |
| Open license | Available | Apache-2.0 model license. |
| Training code | Unknown | The code and complete training recipe for this model have not been verified. |
| Training data | Unknown | A complete public training corpus or reconstruction recipe has not been verified. |
| Paper / report | Available | A technical report describes this model family. |
| Model documentation | Available | Public model card. |
Reviewed 2026-10-13 · Link to this model
Qwen/Qwen2-VL-7B-Instruct
| Check | Status | Evidence |
|---|---|---|
| Weights | Available | Downloadable checkpoint files. |
| Open license | Available | Apache-2.0 model license. |
| Training code | Unknown | The code and complete training recipe for this model have not been verified. |
| Training data | Unknown | A complete public training corpus or reconstruction recipe has not been verified. |
| Paper / report | Available | A technical report describes this model family. |
| Model documentation | Available | Public model card. |
Reviewed 2026-10-13 · Link to this model
Tesseract/tessdata_best-dan
| Check | Status | Evidence |
|---|---|---|
| Weights | Available | The traineddata file is publicly downloadable. |
| Open license | Available | Apache-2.0 model repository license. |
| Training code | Partial | Tesseract training tools are public. The exact recipe for this shipped language file has not been verified. |
| Training data | Partial | Language training texts are released; the langdata repository notes that some training fonts are commercial. |
| Paper / report | Unknown | A technical paper about this model has not been verified. |
| Model documentation | Available | The model repository documents the files and their use. |
Reviewed 2026-10-13 · Link to this model
Tesseract/tessdata_best-nor
| Check | Status | Evidence |
|---|---|---|
| Weights | Available | The traineddata file is publicly downloadable. |
| Open license | Available | Apache-2.0 model repository license. |
| Training code | Partial | Tesseract training tools are public. The exact recipe for this shipped language file has not been verified. |
| Training data | Partial | Language training texts are released; the langdata repository notes that some training fonts are commercial. |
| Paper / report | Unknown | A technical paper about this model has not been verified. |
| Model documentation | Available | The model repository documents the files and their use. |
Reviewed 2026-10-13 · Link to this model
Tesseract/tessdata_best-swe
| Check | Status | Evidence |
|---|---|---|
| Weights | Available | The traineddata file is publicly downloadable. |
| Open license | Available | Apache-2.0 model repository license. |
| Training code | Partial | Tesseract training tools are public. The exact recipe for this shipped language file has not been verified. |
| Training data | Partial | Language training texts are released; the langdata repository notes that some training fonts are commercial. |
| Paper / report | Unknown | A technical paper about this model has not been verified. |
| Model documentation | Available | The model repository documents the files and their use. |
Reviewed 2026-10-13 · Link to this model
HuggingFaceTB/SmolVLM2-2.2B-Instruct
| Check | Status | Evidence |
|---|---|---|
| Weights | Available | Downloadable checkpoint files. |
| Open license | Available | Apache-2.0 model license. |
| Training code | Partial | The official repository contains VLM training code. The exact recipe for this 2.2B checkpoint has not been verified. |
| Training data | Partial | The card lists public dataset sources and mixture proportions. Availability of every selected training sample has not been verified. |
| Paper / report | Available | A technical report describes this model family. |
| Model documentation | Available | Public model card. |
Reviewed 2026-10-13 · Link to this model
google/gemma-4-E2B-it
| Check | Status | Evidence |
|---|---|---|
| Weights | Available | Downloadable checkpoint files. |
| Open license | Available | Apache-2.0 model license. |
| Training code | Unknown | The code and complete training recipe for this model have not been verified. |
| Training data | Unknown | A complete public training corpus or reconstruction recipe has not been verified. |
| Paper / report | Available | A technical report describes this model family. |
| Model documentation | Available | Public model card. |
Reviewed 2026-10-13 · Link to this model
google/gemma-3-4b-it
| Check | Status | Evidence |
|---|---|---|
| Weights | Available | Weights are available after accepting the model terms. |
| Open license | Not open | The custom Gemma terms include use restrictions; this is not an open license under these criteria. |
| Training code | Unknown | The code and complete training recipe for this model have not been verified. |
| Training data | Unknown | A complete public training corpus or reconstruction recipe has not been verified. |
| Paper / report | Available | A technical report describes this model family. |
| Model documentation | Available | Google publishes a model card. |
Reviewed 2026-10-13 · Link to this model
Qwen/Qwen3.5-2B
| Check | Status | Evidence |
|---|---|---|
| Weights | Available | Downloadable checkpoint files. |
| Open license | Available | Apache-2.0 model license. |
| Training code | Unknown | The code and complete training recipe for this model have not been verified. |
| Training data | Unknown | A complete public training corpus or reconstruction recipe has not been verified. |
| Paper / report | Unknown | A technical paper about this model has not been verified. |
| Model documentation | Available | Public model card. |
Reviewed 2026-10-13 · Link to this model
datalab-to/chandra-ocr-2
| Check | Status | Evidence |
|---|---|---|
| Weights | Available | Downloadable checkpoint files. |
| Open license | Not open | The modified OpenRAIL-M terms impose commercial and competing-service restrictions, so this does not meet the open-license check. |
| Training code | Unknown | The code and complete training recipe for this model have not been verified. |
| Training data | Unknown | A complete public training corpus or reconstruction recipe has not been verified. |
| Paper / report | Unknown | A technical paper about this model has not been verified. |
| Model documentation | Available | Public model card. |
Reviewed 2026-10-13 · Link to this model
Qwen/Qwen3.5-4B
| Check | Status | Evidence |
|---|---|---|
| Weights | Available | Downloadable checkpoint files. |
| Open license | Available | Apache-2.0 model license. |
| Training code | Unknown | The code and complete training recipe for this model have not been verified. |
| Training data | Unknown | A complete public training corpus or reconstruction recipe has not been verified. |
| Paper / report | Unknown | A technical paper about this model has not been verified. |
| Model documentation | Available | Public model card. |
Reviewed 2026-10-13 · Link to this model
google/gemma-4-E4B-it
| Check | Status | Evidence |
|---|---|---|
| Weights | Available | Downloadable checkpoint files. |
| Open license | Available | Apache-2.0 model license. |
| Training code | Unknown | The code and complete training recipe for this model have not been verified. |
| Training data | Unknown | A complete public training corpus or reconstruction recipe has not been verified. |
| Paper / report | Available | A technical report describes this model family. |
| Model documentation | Available | Public model card. |
Reviewed 2026-10-13 · Link to this model
Mistral/mistral-ocr-4-1
| Check | Status | Evidence |
|---|---|---|
| Weights | Not open | Distributed as a hosted service, without a public model-weight download. |
| Open license | Not open | API access does not provide an open license to model weights. |
| Training code | Unknown | The code and complete training recipe for this model have not been verified. |
| Training data | Unknown | A complete public training corpus or reconstruction recipe has not been verified. |
| Paper / report | Unknown | A technical paper about this model has not been verified. |
| Model documentation | Available | Public model documentation describes capabilities and API usage. |
Reviewed 2026-10-13 · Link to this model
OpenAI/gpt-6.1-sol
| Check | Status | Evidence |
|---|---|---|
| Weights | Not open | Distributed as a hosted service, without a public model-weight download. |
| Open license | Not open | API access does not provide an open license to model weights. |
| Training code | Unknown | The code and complete training recipe for this model have not been verified. |
| Training data | Unknown | A complete public training corpus or reconstruction recipe has not been verified. |
| Paper / report | Unknown | A technical paper about this model has not been verified. |
| Model documentation | Available | Public model documentation describes capabilities and API usage. |
Reviewed 2026-10-13 · Link to this model