Up to date
baidu Apache 2.0 out 7 months ago
4.74B parameters, images and text in, text out, 32K tokens of context.
Verified 2026-10-25 Rank #296 in LLMs · score 84.1 method v1 Sources 3 Downloads 30d 42.3k · #395 of 619 tracked
Picked by rule from 8 tracked pieces: the lab's own post, the most shared, a field report, the latest.
The 1 pick above, plus these 7. Top is the radar's score: discussion, source authority and the lab's own posts, fading with age.
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: 5 quantized builds and 15 Spaces.
Top Spaces running it: baidu/Qianfan-OCR-Demo ♥ 44 · jason1966/Qianfan-OCR-MLX ♥ 17
Weights only, an estimate: leave room for the KV cache. Method
About 2.8 GB at INT4, 5.7 GB at FP8, 11 GB at BF16: weights only, so leave room for the KV cache.
Estimate params × bytes per param × 1.2 · how we estimate
Weights, every dated article, and repo commits on one rail.
Teal is the weights (2026-03-18), amber triangles are 4 dated articles (each links to it), grey ticks are repo commits.
6 derivatives on the Hub. Which build to run, by runtime: the most downloaded in each class.
From the Hugging Face model tree; quantized builds in teal, fine-tunes, merges and adapters in greys.
Everything else, folded: the full tables, every source.
| Benchmark | Value | Setting / notes | Reported by | Date |
|---|---|---|---|---|
| author-reported Terminal-Bench · Write code with agents | 85.9 | — | model author | 2026-10-07 |
| author-reported TextVQA · Understand images and video | 80 | — | model author | 2026-10-07 |
| author-reported olmOCR-bench · Extract text from documents | 79.8 | — | model author | 2026-10-07 |
| author-reported overall (allenai/olmOCR-bench) | 79.8 | — | Qianfan-OCR technical report | 2026-10-24 |
| author-reported arxiv_math (allenai/olmOCR-bench) | 80.1 | — | Qianfan-OCR technical report | 2026-10-24 |
| author-reported old_scans_math (allenai/olmOCR-bench) | 73.1 | — | Qianfan-OCR technical report | 2026-10-24 |
| author-reported table_tests (allenai/olmOCR-bench) | 81.6 | — | Qianfan-OCR technical report | 2026-10-24 |
| author-reported old_scans (allenai/olmOCR-bench) | 42 | — | Qianfan-OCR technical report | 2026-10-24 |
| author-reported multi_column (allenai/olmOCR-bench) | 80.4 | — | Qianfan-OCR technical report | 2026-10-24 |
| author-reported long_tiny_text (allenai/olmOCR-bench) | 89.1 | — | Qianfan-OCR technical report | 2026-10-24 |
| author-reported headers_footers (allenai/olmOCR-bench) | 92.2 | — | Qianfan-OCR technical report | 2026-10-24 |
| author-reported baseline (allenai/olmOCR-bench) | 99.6 | — | Qianfan-OCR technical report | 2026-10-24 |
The multilingual model performs direct image-to-Markdown conversion and supports tasks like table extraction and document question answering.[1][2] It is released under the Apache 2.0 license and utilizes a vision-language architecture.[3][4]
Generated 2026-10-07 15:05 UTC by parasail-gemma-4-26b-a4b-it; each numeral points at the fact it rests on. Not a verdict.
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.
Listed in LLMs
| Pipeline tag | image-text-to-text |
|---|---|
| Library | transformers |
| Architectures | QianfanOCRForConditionalGeneration |
| Model type | qianfan_ocr |
| Dtype | bfloat16 |
| Params by dtype | BF16: 4,741,408,256 |
| Size band | M |
| MoE | no |
| Context length | 32,768 (config.json max_position_embeddings) |
| Languages | multilingual |
| Papers | arXiv:2603.13398, arXiv:2509.18189 |
| License | Apache 2.0 (apache-2.0) |
| Hub snapshot | 2026-10-25 |
vLLM: supported · QianfanOCRForConditionalGeneration · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: not in registry · 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.
Weights-only memory for Qianfan-OCR: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.
| BF16 (2 bytes/param) | 11 GB · Single GPU (<24 GB) |
|---|---|
| FP8 (1 byte/param) | 5.7 GB · Single GPU (<24 GB) |
| INT4 (0.5 bytes/param) | 2.8 GB · Edge (<4 GB) |
Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra.
quantized: 5finetune: 1
| Variant | Relation | Quant | Params | Band | License | Author | Likes | Downloads |
|---|---|---|---|---|---|---|---|---|
| Reza2kn/Qianfan-OCR-GGUF | quantized | GGUF | — | — | apache-2.0 | Reza2kn | 16 | 19.6k |
| Abiray/Qianfan-OCR-GGUF | quantized | GGUF | — | — | apache-2.0 | Abiray | 1 | 4.3k |
| jason1966/Qianfan-OCR-MLX-4bit | quantized | MLX | 4.74B | M | apache-2.0 | jason1966 | 5 | 2.4k |
| DevQuasar/baidu.Qianfan-OCR-GGUF | quantized | GGUF | — | — | — | DevQuasar | 0 | 2.1k |
| bssrdf/Qianfan-OCR-gguf | quantized | gguf | — | — | apache-2.0 | bssrdf | 0 | 474 |
| mat50013/Qianfan-OCR-MLX-BF16 | finetune | MLX | 4.74B | M | apache-2.0 | mat50013 | 0 | 409 |
| Date | Commit | SHA |
|---|---|---|
| 2026-04-29 | Upload folder using huggingface_hub | 623bf5d |
| 2026-03-26 | Update README.md | 5e181e7 |
| 2026-03-19 | correct example in readme (#6) | 17fd115 |
| 2026-03-18 | Update README.md | 69ccfc5 |
| 2026-03-18 | Upload folder using huggingface_hub | 6a53ad3 |
| 2026-03-18 | Add OlmOCRBench evaluation results (#1) | f0d830f |
| 2026-03-18 | Upload folder using huggingface_hub | e2b569a |
| 2026-03-18 | initial commit | 0b0901d |
Ranked by source signals (HN, Lobsters, authority domains, official posts) and recency. Summaries are generated from a stored copy of the article text; every sentence cites a span of that text, and nothing is shown that failed the citation check.
arXiv links found in the model card and its tags. "Peer-reviewed" is shown only with evidence: a published venue in OpenAlex (measured) or the authors' own journal reference on arXiv (reported).
Hub snapshot: 2026-10-25 · HF API: 2026-10-24 21:37 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:45 UTC · Summary: 2026-10-07 15:05 UTC · Last verified 2026-10-25
Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.
