There's the NCI Genomic Data Commons:
For all submitted sequence data, including BAM alignment files, the GDC generates new alignments in BAM format using the latest human reference genome GRCh38 with standard alignment pipelines. Using these standard alignments, the GDC generates high level derived data, including normal and tumor variant and mutation calls in VCF and MAF formats, and gene and miRNA expression and splice junction quantification data in TSV formats.
But that needs an NIH account for individual-level data, which might be too restrictive for you.
There's also the Texas Cancer Research Biobank (7 human cancer patients with matched normal):
These tumor and normal specimen sequence files (FASTQ and BAM), somatic variant calls (VCF and MAF) and germline MAF files are available for each consented patient as described on the TCRB Open Access Privacy page.
That also requires account registration, but the registration process is automated and much simpler (requires just a name, and acceptance of usage conditions). There's a published paper about that dataset.
I had some vague feeling that another repository existed (created within the last year), but I can't seem to find it from quick web searches.