Latest open-source achievements (#21): A feast of open-source models! Gemma 4, DeepSeek V4, Kimi K2.6, MiMo 2.5, GLM-5.1, etc. Evaluation based on CAISI V4.

ME News updates: On May 17 (UTC+8), the open-source AI model sector saw a flurry of releases this month, including a series of major new versions such as Gemma 4, DeepSeek V4, Kimi K2.6, MiMo 2.5, and GLM-5.1. These models were all performance-tested under CAISI’s V4 evaluation framework. Multiple teams rolled out flagship-level updates almost simultaneously, reflecting the current fiercely competitive landscape of rapid iteration in the open-source community. (Source: AiHot)
View Original
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
  • Reward
  • 7
  • 1
  • Share
Comment
Add a comment
Add a comment
NfaKitchen
· 4h ago
Everyone is choosing to announce at this time—was it prearranged or just a coincidence?
View OriginalReply0
GateUser-14cb5f72
· 7h ago
The infrastructure behind “intensive release announcements” has matured—only after the training costs are brought down can we play like this.
View OriginalReply0
GateUser-c4e25c95
· 7h ago
Kimi K2.6 has been rumored before its release, and it has finally come to fruition.
View OriginalReply0
GateUser-0b71fc11
· 7h ago
Did GLM-5.1 catch up with the first tier this time? Provide a detailed comparison.
View OriginalReply0
PatchNotePaladin
· 7h ago
MiMo 2.5 passes by quietly, but actual tests seem to show it's quite capable
View OriginalReply0
OwlChainViewer
· 7h ago
Is the CAISI V4 evaluation framework now an industry standard? Everyone is jumping on the bandwagon.
View OriginalReply0
SmallPosition,BigMouth
· 7h ago
The open-source AI craze is getting out of hand—Gemma4 and DeepSeek V4 are released at the same time, and users basically win effortlessly.
View OriginalReply0
  • Pinned