Lightnews — Scholar-powered news

@giyaibo.bsky.social

12 followers 90 following 0 posts

Posts Replies Media Videos

Reposted

Tim Kellogg

@timkellogg.me

DeepSeek-OCR

a tiny 3B-A0.5B MoE OCR model that runs fast on a single A100 40GB with very high precision and excellent compression

why it’s cool — they use images as a way to compress text and get around the O(n^2)

huggingface.co/deepseek-ai/...

A scatter plot titled “Overall Performance (Edit Distance) vs Average Vision Tokens per Image” compares OCR and vision-language models by token efficiency and accuracy.

Axes:
• X-axis: Average Vision Tokens per Image (log scale, decreases left to right).
• Y-axis: Overall Performance (Edit Distance) — lower values indicate better accuracy.

⸻

Color legend (bottom-left):
• 🔴 DeepEncoder Series
• 🟩 QwenEncoder Series
• 🔵 InternVLEncoder Series
• 🟧 Other Encoders

⸻

Highlighted regions:
• Left (purple box): “Vision Tokens > 1500, Average per image (← More)”
• Right (blue box): “Vision Tokens < 1000, Average per image (→ Fewer)”
• Green box: “High Accuracy ED < 0.25 (↑ better)”

⸻

Key models:

DeepEncoder Series (red circles):
• DeepSeek-OCR (Large, Base, Small, Tiny, Gundam, Gundam-M 200dpi) — clustered near the top-right with high accuracy (≈0.1–0.25 ED).
• DeepSeek-OCR (Gundam-M 200dpi) achieves the best performance.

QwenEncoder Series (green squares):
• dots.ocr, Qwen2.5-VL-72B, OCRFlux-3B, Qwen2.5-VL-7B, OLMOCR — around mid-range (0.25–0.4 ED) with 1000–5000 tokens per image.
• dots.ocr (200dpi) is among the top in this group.

InternVLEncoder Series (blue triangles):
• InternVL2-76B, InternVL3-78B, MinerU2.0 — higher token usage (4000–7000) with moderate accuracy (0.2–0.45 ED).

Other Encoders (orange diamonds):
• GOT-OCR2.0 (mid performance)
• SmolDocling (bottom-right, 400 tokens/image, lowest accuracy ≈0.5 ED).

⸻

Summary:

Models using fewer vision tokens (right side) generally have worse accuracy, while those with more tokens per image (left side) perform better.
DeepSeek-OCR (Gundam-M 200dpi) leads overall in accuracy, while SmolDocling is the smallest and least accurate.

October 20, 2025 at 11:12 AM

Reposted

Simon Willison

@simonwillison.net

Mistral shared what look like the most detailed numbers yet for the environmental impact of training a frontier LLM - their Mistral Large 2 used 20,4 ktCO₂e and 281,000 m3 of water

I'd love to see numbers like that provided in context though, hard to evaluate alone simonwillison.net/2025/Jul/22/...