ToolDriftRE-COUNTED EVERY NIGHT
COUNTED 08:54 UTC4 READS FAILEDRead API
Menu
COUNTED 08:54 UTC4 READS FAILED
60 models ranked12 boards36 tools tracked, 7 not what they werecaptured 27 Sep 2026against 26 Sep 202611 rose · 12 fell · 84 held · 4 newadoption, not capability
CHANGES/THE BOARD/LEADER CHANGE

Best coding model, Under $1 / M: deepseek/deepseek-v4-flash-0731 overtook tencent/hy3

The top-ranked model for Best coding model, Under $1 / M changed from tencent/hy3 to deepseek/deepseek-v4-flash-0731, measured on token volume through OpenRouter.

model id tencent/hy3model id deepseek/deepseek-v4-flash-0731
OCCURRED
11 August 2026
OBSERVED
11 August 2026
SEVERITY
MEDIUM
CONFIDENCE
HIGH
LEADER CHANGEENTRY 188SOURCE auto
Contents
WHAT CHANGEDwas, and is now
WASmodel id tencent/hy3
IS NOWmodel id deepseek/deepseek-v4-flash-0731
WHAT IT MEANSThe top-ranked model for Best coding model, Under $1 / M changed from tencent/hy3 to deepseek/deepseek-v4-flash-0731, measured on token volume through OpenRouter.
OCCURRED11 August 2026the day the change took effect, where the vendor said so
OBSERVED2026-08-11 09:51:12 UTCthe night this site read it and saw a difference
EVIDENCEopenrouter.ai
HOW THIS WAS FOUNDhigh confidence

Derived from an API, a repository status or an ownership record, not a reading of a marketing page. Source: auto. Medium severity says how much it changes what the tool costs or what you get free.

EVERY BOARD, IN FULLALL CHANGES