Open-weight OCR got so cheap I had to share it
Details
- External ID
- 49041317
- Source
- HN
- Company
- —
- Product
- Open-weight OCR got so cheap I had to share it
- Website domain
- openparser.dev
- Launched
- July 24, 2026
- Cohort
- —
- Upvotes
- 17
- Upvotes percentile
- 0.6726403823178017
- Tags
- —
- Fetched at
- Sept. 7, 2026, 9:26 p.m.
- Updated at
- Sept. 7, 2026, 9:26 p.m.
Description
This was not supposed to become a product.When PaddleOCR-VL-1.6 dropped, independent benchmarks put it at the top of document parsing models. I had to try it. I needed a provider, but there simply isn't one ready for production that I would trust.So i set one up myself. I assumed that even after getting it running, serving a vision-language model would be expensive.It turns out the opposite is true. Once I had it running properly, the cost was absurdly low. At proper GPU utilization, the cost is only around $1 per 1,000 pages.The nearest competitors are either much lower quality (Azure Read) or absurdly expensive (Extend or Reducto). Even Mistral OCR 4 which is really good and pretty cheap is still 4x more expensive.So I had to share it. I made the endpoitns public, and vibe-coded a simple dashboard.Let me know what you think!
Enrichment
- Theme
- lightweight and on-device AI runtimes
- Vertical
- Horizontal
- Function
- Dev tools
- Audience
- Developer
- AI stance
- Not AI
- Project type
- Hobby / open-source project
- Normalized one-liner
- open-weight ocr model
- Manually corrected
- False
Could you build this?
Partial Setting up a simple wrapper around PaddleOCR is easy, but running an ultra-cheap, highly performant hosted OCR API with autoscaling GPU infrastructure, multi-tenant queues, and low-cost batch processing requires substantial systems and infrastructure expertise.
What it would actually take: A production deployment needs an orchestration layer (e.g., Ray Serve, Triton Inference Server, or custom vLLM/vLLM-like pipelines) hosted across GPU spot instances (RunPod, Lambda, AWS). It requires robust asynchronous request queueing (Redis/SQS), memory-optimized model inference, PDF page splitting/rendering pipelines, and complex load-balancing to keep inference costs low enough to support $1/1,000 pages.
Discussion
11 comments analyzed.
Competitors mentioned: PaddleOCR, AWS, OpenAI, OpenRouter
Concerns raised: Pricing not justified by openness, just GPU costs, Misleading 'open' branding without self-hosting documentation, Vendor lock-in despite open-weight claims, Monetizing open-source model as SaaS without clear transparency
Feature requests: Self-hosting documentation, Better handwritten text parsing
Competitors
Other products that read as similar to this one — 50 launches clear the similarity bar, closest 8 shown.
Attention rank: #23 of 51 (itself plus its competitors, highest first — normalized so YC and Product Hunt are compared fairly).
Launched 266 days after the earliest competitor.
- Hosted PaddleOCR-VL-1.6 API · hn · 2026-07-23 · 5 upvotes · similarity 0.80
- Run open-weight OCR, VLM and vision models behind one API · hn · 2026-09-04 · 5 upvotes · similarity 0.49
- DocsRouter · hn · 2025-12-18 · 14 upvotes · similarity 0.47
- OCR Arena · hn · 2025-11-21 · 216 upvotes · similarity 0.43
- We benchmarked 18 LLMs on OCR (7K+ calls) · hn · 2026-04-22 · 5 upvotes · similarity 0.42
- Moonshine Open-Weights STT models · hn · 2026-02-24 · 316 upvotes · similarity 0.42
- Online OCR Free · hn · 2026-03-03 · 14 upvotes · similarity 0.41
- DeepSeek Flash inverted the economics of agent products · hn · 2026-06-25 · 9 upvotes · similarity 0.41
Other launches for this product
- No other launches for this product.
Same idea, different domain
Nobody's really built a dev tools tool for Sales yet.