The Open Weights
LatestModelsLeaderboardsCompanies
Subscribe
The Open Weights

The daily record of open-source AI. New model releases, leaderboards, and what's coming next — written for people who ship.

Refreshed every 12 hours

Discover

  • Latest releases
  • New today
  • Trending models

Browse

  • All models
  • Companies
  • Categories
  • Leaderboards

About

  • About
  • Editorial policy
  • Hardware estimates
  • RSS feed
  • llms.txt
  • Newsletter

© 2026 The Open Weights. An independent publication.

PrivacyTermsSMSAggregated by Claude · curated by humans.
LatestXingChen AGI1
XingChen AGIVision-Language

TeleOCR brings bilingual document parsing to Qwen2.5-VL

A new vision-language model targets Chinese and English OCR and structured document extraction, released openly on Hugging Face.

Aug 14, 2026
NotableOther
TeleOCR

TeleOCR, a new vision-language model focused on optical character recognition and document parsing, has arrived on Hugging Face. Built on top of Alibaba's Qwen2.5-VL foundation, it targets both Chinese and English text, positioning itself for the messy reality of real-world documents rather than clean, single-language scans.

Document understanding has become one of the more practical proving grounds for open vision-language models. Beyond simply reading characters, the task increasingly means preserving layout, tables, and reading order so that downstream systems can turn a scanned page into usable structured data. TeleOCR is pitched at exactly this class of work.

Why it matters

Bilingual OCR is harder than it looks. Chinese-English mixed documents — invoices, forms, reports, and contracts — routinely trip up models tuned for a single script. By fine-tuning a capable general VLM like Qwen2.5-VL for this purpose, TeleOCR aims to give developers an open alternative to closed OCR APIs.

A few things to keep in mind about this first release:

  • It is an initial version (v1) built on the Qwen2.5-VL base.
  • The license is listed as "other," so teams should review terms before commercial use.
  • Parameter count, context length, and benchmark figures were not specified in the release record.

As with any early open release, the real test will be how TeleOCR performs against established OCR pipelines on complex layouts. For now, the model is available to download and evaluate directly from its Hugging Face repository.

Sources

  • XingChen-AGI/TeleOCR

    Hugging Face

    Visit
OlderTencent's UI-Mate-27B targets desktop automationTencent · Vision-Language · 3 months agoNewerFastino releases GLiNER 2.5 Multi for extractionFastino · Text / LLM · 3 months ago

Get the model

Hugging Face

Specs

Context window128K tokens
Size2.8 GB
PrecisionBF16
ArchitectureQwen2_5_VLForConditionalGeneration
LicenseOTHER
Downloads18.2K
Likes1.3K

Can you run it?

Runs on just about anything.

  • BF16 (as published)

    8 GB GPU (RTX 4060) · 16 GB laptop or Mac

    4.7 GB
  • 8-bit

    Any modern laptop · 8 GB RAM

    3.4 GB
  • 4-bit

    Any modern laptop · 8 GB RAM

    2.8 GB
Your machine
GGUF builds
  • KV cache sized for 8,192 tokens of context; longer prompts need proportionally more.

Based on Hugging Face weights metadata. Assumes an 8K context and ~1 GB runtime overhead; actual needs vary. How we estimate


Modalities

Vision-Language
2 versions — view changelog

The Weekly Weights

Every open release that mattered, one email a week.

0 comments

No comments yet. Be the first to weigh in.

More from XingChen AGI

All XingChen AGI releases →
Xing4.0-29B-A4B
XingChen AGI/Text / LLMConsumer GPU

Xing4.0 arrives as a 29B MoE with 4B active params

inclusionAI's new text model uses a mixture-of-experts design to keep compute low while shipping under an Apache-2.0 license.

Sep 16, 2026

More in Vision-Language

All Vision-Language →
SenseNova-Loop-9B
SenseTime/Vision-LanguageRuns on a laptop

SenseTime Releases Looped 9B Multimodal Model

The company's new SenseNova-Loop-9B pairs vision-language capabilities with a recurrent looped-transformer design.

Nov 17, 2026
Hunyuan Hy4-flash
Tencent/Vision-LanguageMulti-GPU server

Tencent releases Hunyuan Hy4-flash multimodal MoE

The lightweight, mixture-of-experts model adds vision and text instruction-following to Tencent's open Hunyuan lineup.

Nov 17, 2026
Mirage
Cactus Compute/Vision-Language

Cactus Compute's Mirage brings vision tasks on-device

The new open vision-language model targets detection, segmentation, and visual Q&A without relying on the cloud.

Nov 5, 2026