Assigning condition and coldata

condition <- factor(c("C","C","C","P","P","P"))
coldata <-data.frame(row.names = colnames(counts_filtered)[-1], condition)

#Deseq
dds<-DESeqDataSetFromMatrix(countData = counts_filtered[, -1],, colData = coldata, design = ~condition)
dds <- DESeq(dds)

#extracting deseq2 results
res <- results(dds, contrast = c ("condition", "C","P"))
res

above is the code I am using to run deseq2. When I look at the results they are different, upregulated genes according to the raw read counts showing me a negative value and the genes which pretty much looks like downregulated in the read counts showing me a positive value. This is between three cases and 3 controls. For example, the read counts of one gene across cases and controls:

C01 C02 C03     P01    P02    P03
14  47  19      129874  25114  31346

and desq results for this gene is as follows:

baseMean                 log2FoldChange

23071.1919508785    -10.843407253606

Can anybody help with this?

I don't see why this is unrealistic. A logFC of -10 is about -1000 on normal scale. The rough difference between the mean of these counts per group (are they normalized?) as roughly in this range. Your contrast is C - P so negative logFCs mean it's higher in P, so this is all in line.