Gemini-3.8-Flash

Headlines · 3

1. 3d
   Elorian launches Glimpse and DecoyCount benchmark

   Elorian launched its first visual reasoning model, Glimpse, and co-released the DecoyCount benchmark with Mercor. The benchmark tests whether models genuinely count objects or rely on visual shortcuts and substitution heuristics. According to the accompanying paper, all evaluated frontier models scored below 50%, with GPT-6 Astra highest at 43.17%. Elorian CEO Andrew Dai said Glimpse scored 50.48% on the benchmark’s 1,660 items in an internal test that was not included in the paper. He said the evaluation images and text were withheld from Glimpse’s training and that a contamination check found no overlap in the data checked. Dai also said Glimpse’s result came from a single run without an agent or tool harness, while the six models in the paper were averaged across three runs. Elorian describes Glimpse as a visual thinking system that generates and edits images as part of its reasoning process, distinguishing the company’s approach from world models.

   Latent Space · 🔥 0
2. Oct 19
   S-Agent 2.0 gives a general coding agent a perception toolkit for spatial reasoning

   An arXiv paper submitted on 20 October 2026 by Rui Gong and 11 co-authors presents S-Agent 2.0, a spatial reasoning agent that equips an off-the-shelf general coding agent with a command-line perception toolkit and a persistent scene memory, so the model acquires scene evidence, remembers it and interprets readings in plain spatial language. The authors report the toolkit lifts Qwen3.8-27B by 13.9 points on MMSI-Bench and 20.5 points on ReVSI over the same agent without tools, and Gemini-3.8-Flash by 11.2 and 9.6 points; with the open backbone S-Agent 2.0 scores 66.5 and 82.3, ahead of SpatialClaw on the same backbone (62.6 and 74.5), and with Gemini-3.8-Flash it reaches 74.1 and 90.5 against 67.0 and 79.7. The paper says the system also works across different coding-agent harnesses such as Codex and Claude Code, and that tool gains shrink as the backbone becomes more capable, though the authors say their gain on ReVSI holds at 9.6 points while SpatialClaw's falls from 13.2 to 2.7.

   Hugging Face · Papers · 🔥 0
3. Oct 18
   VAST open-sources EdenLive for real-time generative rendering

   VAST has released and open-sourced EdenLive, the real-time generative renderer within its Project Eden research program, according to 机器之心, 爱范儿 and 量子位. It uses proxy scenes or reference videos, a target first frame and text prompts to render changing visuals while an upstream system maintains interaction rules and world state. VAST says the 5B model exceeds 24 FPS at 720P on one Nvidia RTX 5090 and supports minute-scale continuous generation; the release also includes 5B and 14B weights, streaming inference code and technical documentation.

   爱范儿 · 🔥 0

Experience and discussion from the community

Share my Gemini-3.8-Flash experienceAsk about Gemini-3.8-Flash