Sonnet 5.5 vs GPT-6.1 Sol for video scripts: both $2 in, $10 out

If a script writer step feeds a video agent, the LLM bill is rarely the large number, and the two newest mid-tier models list identical prices. Anthropic's models overview lists Claude Sonnet 5.5 (claude-sonnet-5-5) at $2 per million input tokens and $10 per million output tokens, with a 1M-token context window and 128K max output. OpenAI's models page lists GPT-6.1 Sol (gpt-6.1-sol) at $2 and $10 with a 1.05M context window and 128K max output (both read 2026-10-05).

What a script call costs

A 600-word voiceover script is roughly 800 output tokens. At $10 per million that is $0.008, under one cent, with the same figure on either model. Even with a 20,000-token brief as input ($0.04), the writing step stays far below one generated clip. The decision is therefore about whether the model follows your structure, not the rate card.

Where Sume's side differs

Sume's Agent Completions take an instruction or messages[] and require generation_spend_cap_usd, which bounds generation spend (clips, images, voice), not tokens. The Scheduled object carries a model field alongside its instructions, cron expression and cap. The docs do not list which LLMs the agent can use, so check the dashboard before you plan a model-by-model test.

Run the same brief through both for a week, keep the one whose scripts need fewer edits, and re-read both vendor pages first: prices and ids moved more than once this year.