VQV Signal

RESEARCH WATCH TECHNICAL

Gemini 4 Neon tops LlamaIndex's ParseBench-2 document benchmark, per Jerry Liu

According to a post on X by @jerryjliu0, LlamaIndex's Jerry Liu says Neon, running on low thinking, beat recent models from Opus 5.5 to GPT-6.2 Sol at parsing complex tables and placing bounding boxes. The post notes that at about 6.2 cents a page, Neon is expensive. It also says Google called Neon its third Gemini model in eight days, and T3 Code's Theo says Gemini usage in his tool is growing exponentially.

Source: Scobleizer News Pulse (X) · x.com Found via Robert Scoble’s Scobleizer News Pulse · a post on X, not independently verified Published 2026-11-21T02:45:00+00:00 Detected 2026-11-21T15:01:37+00:00
View original source

According to a post on X by @jerryjliu0, LlamaIndex's Jerry Liu says Neon, running on low thinking, beat recent models from Opus 5.5 to GPT-6.2 Sol at parsing complex tables and placing bounding boxes. The post notes that at about 6.2 cents a page, Neon is expensive. It also says Google called Neon its third Gemini model in eight days, and T3 Code's Theo says Gemini usage in his tool is growing exponentially.

AI-assisted summary based on the listed source.

LlamaIndex's Jerry Liu says Neon, on low thinking, beat recent models from Opus 5.5 to GPT-6.2 Sol at parsing complex tables and placing bounding boxes, though at about 6.2 cents a page it is expensive. Google called Neon its third Gemini model in eight days, and T3 Code's Theo says Gemini usage in his tool is...

The source describes document parsing as high-volume enterprise work, so table and layout accuracy at scale could matter to builders and businesses that process many documents. Cost is the trade-off to watch, since the post itself calls Neon expensive per page. This is a claim from one social post, and it is not an independent verification of the benchmark result.

Signal Strength 88% Technical label WATCH Public Interest 37 Category RESEARCH Reader Depth TECHNICAL

Signal Strength reflects source quality, relevance, freshness and evidence. Public Interest helps organize discovery; it is not proof of truth.

Public Interest components
Recognizable Entity Score 0 Practical Impact Score 28 Novelty Interest Score 94 Consequence Score 50 Curiosity Score 0 Shareability Score 50

VQV surfaced this signal because it is recent, relevant to Open Source LLMs, connected to Scobleizer News Pulse (X).