DeepSeek V4.1 Flash

connectable

Open-weight model we do not host ourselves. You can still use it through the privacy layer.

97 Logic Index · via OpenRouter
– tokens per second here
– no tier here

Fact sheet

Developer
DeepSeek
Size
763B parameters
Context
1M
Input
Text, Image
License
MIT
Hosting
not hosted here, connectable via the privacy layer

Call

Connectable: open-weight models we do not host ourselves, for example because they are too large for our hardware. You can still use them through our privacy layer: personal data is replaced with placeholders before the request leaves. Measured via OpenRouter with the same test.

By test area

Reasoning chains
98
Logic vs. world knowledge
100
Logic grid puzzles
94

Consistency: same answer in every run for 99% of tasks.

More models