deepseek-r1:14b

57.3M 1 month ago

DeepSeek-R1 is a family of open reasoning models with performance approaching that of leading models, such as O3 and Gemini 2.5 Pro.

tools thinking 1.5b 7b 8b 14b 32b 70b 671b
... /
template
58174d17c07a · 537B
{{- if .System }}{{ .System }}{{ end }}
{{- range $i, $_ := .Messages }}
{{- $last := eq (len (slice $.Messages $i)) 1}}
{{- if eq .Role "user" }}<｜User｜>{{ .Content }}
{{- else if eq .Role "assistant" }}<｜Assistant｜>
{{- if and $last .Thinking -}}
<think>
{{ .Thinking }}
</think>
{{- end }}{{ .Content }}{{- if not $last }}<｜end▁of▁sentence｜>{{- end }}
{{- end }}
{{- if and $last (ne .Role "assistant") }}<｜Assistant｜>
{{- if and $.IsThinkSet (not $.Think) -}}
<think>
</think>
{{ end }}
{{- end -}}
{{- end }}