Unverified50% confidenceFactExact time
In a video simulation test of double-pendulum chaotic motion, the 1017 version tracked the light points on the pendulum arms across all 500 frames and fitted the physical parameters, whereas Gemini's Argon measured the pendulum angles and pulled them back with a hidden spring.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/19/2026
First Seen
Valid until: 1/17/2027
Sources
DeepSeek V4.1 Flash 1017 Hands-On: Token Efficiency Fixes Its Biggest Weakness
youtubePrompt Engineering10/19/2026
Related Entities
Related Claims
UnverifiedThe 60 FPS Gemini 4 Argon self-reported in the 3D blackjack project came from a counter capped at 60 and was measured in a small headless window, with no rule tests at all.68% similarUnverifiedFrom a single image of a Gothic Victorian carriage, Gemini 4 Argon generated an interactive 3D model with 174,000 faces entirely through code, including openable doors, physical wheel effects, a suspension system, flickering lanterns and skull carvings, without calling Blender.68% similarUnverifiedIn the subway FPS test, Gemini 4 Argon's first-round scene was rather simple, and after one round of "AAA-level graphics" instructions the enemy models, environment textures and bullet hole effects improved significantly; the detail of the carriage hand straps swaying with the train had previously appeared only in Meta Muse Wonder.66% similarUnverifiedIn testing, Gemini 4 Argon was asked to rebuild a screenshot of a landing page nearly 6000 pixels tall using pure static HTML and CSS, and it measured via self-written scripts, rendered with a headless browser, generated pixel-level difference heatmaps, and self-corrected over six consecutive rounds, ultimately achieving a nearly identical result.65% similarUnverifiedThe Gemini 4 Argon team, using a single human demonstration video and the Mojoku simulator, enabled a humanoid robot to learn the cornhole throwing task without preset programming.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1069661API
curl https://kongchang.com/api/v1/knowledge/claims/1069661MCP
get_claim(id=1069661)