Qwen3.8 Max Prime Listed: $4 Input, $12 Output, 1M-Token Context

What Qwen3.8 Max Prime Is

Qwen quietly listed Qwen3.8 Max Prime on 2026-09-23 — an open-source multimodal model accepting text, image, and video inputs and returning text. The headline spec is a 1,000,000-token (1000k) context window, which puts it in contention for long-document, long-video, or multi-source retrieval tasks where context length is the bottleneck.

Pricing in Plain Terms

At $4.00 per 1M input tokens and $12.00 per 1M output tokens, the cost structure follows the now-standard pattern of output being 3× more expensive than input. That ratio matters: if your workload is output-heavy (summarisation, generation, code completion), costs climb faster than a flat per-call rate would suggest.

For comparison, a job that pushes 10M input tokens and produces 2M output tokens would run roughly $64 total — not cheap, but potentially justified if the 1M-token context genuinely replaces chunking pipelines that add latency and complexity.

Who Should Pay Attention

• Long-context video analysis: native video input plus 1M tokens is a meaningful combination for teams processing lengthy recordings.
• Multimodal RAG: combining image and text sources in a single context pass avoids the overhead of separate embedding pipelines.
• Cost-sensitive shops: if $12 output pricing feels steep, check alternatives on our site — several open-weight models in this tier run at lower output rates with comparable context.

Being open-source is worth noting: self-hosting eliminates per-token costs entirely if your infrastructure can handle the model weight, making the listed API price a ceiling rather than a fixed cost.

See Qwen3.8 Max Prime — live specs & price history for up-to-date pricing and side-by-side comparisons.

Ezra, Scout AI Team