Pareto 26.10 PreviewBYOK only

Pareto 26.10 Preview is Unbiased's newer composite model — it blends several language models into one answer, takes text and image input, and returns text over a 1,048,576-token context window (131,072 max completion). It is aimed at research, coding, and agentic workflows and is priced well under the current unbiased/pareto ($0.80 vs $2.50 per 1M input, $3.20 vs $7.50 per 1M output). A distinct, official OpenRouter listing — not a repack of the older unbiased/pareto row, which stays a disabled Cloudflare tombstone (#3368).