$1 Classifies 6,700 Pages: Open-Weight VLMs Read Pixels Directly for Docs
spillai · x · 2026-09-29
typesafeai launched a Jev-compatible API on the vlmrun gateway: open-weight VLMs read pages as pixels (no OCR→LLM hacks) and return one typed, calibrated decision per page at 180ms p50 end-to-end.
The economics are concrete: $1 buys 6,700+ pages, and a full 1,651-page OmniDocBench run cost 25 cents. Supported models include diffusiongemma-26B-A4B-it, gemma4-26B-A4B-it, Qwen3.8-27B and Qwen3.5-0.8B. The team says Jev pushed developers to rethink the cost-pareto front of data processing; free to sign up and try.
More from coding & agent
- Sonnet 5.5 Ships With Opus-Level Writing Gains, But Mid-Tier Models May Be Dying — every · 2026-09-29
- Dev Slams Neon x Vercel Integration for Sneaky $80-100/Mo Preview Branch Charges — FanaHOVA · 2026-09-29
- Meta Muse Falls Short of Codex: No Local Desktop Control, File Upload Bugs — brandon_galang · 2026-09-29
- Burkov: Ugly Code That Works Beats No Code At All — burkov · 2026-09-29
- Scale's Muse agent autonomously files a Verizon complaint from one photo — rickasaurus · 2026-09-29
- Berkeley PhD Candidate Compares Claude Code vs Codex vs Pi: Harness Choice Hits Cost More Than Accuracy — arena · 2026-09-29