DeepSeek launches V4.1 Flash AI model
Photo: South China Morning Post
Ai Added 2d ago 2 outlets

DeepSeek launches V4.1 Flash AI model

Chinese AI developer DeepSeek launched V4.1 Flash, a Mixture-of-Experts model built on a 552 billion-parameter framework that activates 8 billion parameters for input processing and 16 billion for response generation. The company said the model outperformed OpenAI's GPT-5.6 Sol, Moonshot AI's Kimi K3, and its own V4 Pro on the Terminal-Bench 2.1 benchmark with a score of 90.6.

9
Divergence score
2 outlets covered it, splitting into 2 framing camps across 2 bias groups.
2 camps
2 bias groups
Market signalBETA
The spectrum · how 2 outlets placed this story
LeftCenterRight
South China Morning Post
Reuters
Horizontal = outlet biasColor = this story's framing
Supportive of action
Neutral
Dismissive
Critical
Alarmist
International angle
Widest split · the two outlets furthest apart, and who published first
Furthest apart on this story
South China Morning Post
DeepSeek says new Flash AI model beats Kimi K3 on cyber, coding benchmarks
benchmark victory and China AI efficiency race · positive
Reuters
China's DeepSeek launches V4.1-Flash model - Reuters
bare launch announcement · neutral
How South China Morning Post and Reuters compare across every story →
First to report
  1. Reutersfirst
  2. South China Morning Post+2.9 h
By each outlet's earliest publish time in this story. Feeds that index slowly can appear later than they printed.
The split, in one line
SCMP foregrounds beats Kimi K3 on cyber, coding benchmarks and China's chip-curb efficiency race, while Reuters simply launches V4.1-Flash model with no benchmark detail.
How each outlet covered it

No left-right split here

Coverage clusters in the center and international press. Here is each take as it stands.

Center & international coverage
SCSouth China Morning PostINTERNATIONAL2d ago

“DeepSeek says new Flash AI model beats Kimi K3 on cyber, coding benchmarks”

RReutersCENTER2d ago

“China's DeepSeek launches V4.1-Flash model - Reuters”

4 tracked claims across 2 outlets
Fact ledger
All4Claimed3Corroborated1
1/2
Claimed
V4.1 Flash uses a Mixture-of-Experts design built on a 552 billion-parameter framework, activating 8 billion parameters for input and 16 billion for output
Corroborated
Disputed