Same Price Per Token, 7.2x Different Bill: Sonnet 5.5 vs GPT-6 Sol
Claude Sonnet 5.5 (released Sept 28, 2026) and OpenAI's GPT-6 Sol (released Sept 22, 2026) charge identical per-token API pricing — $2 per million input tokens, $10 per million output tokens. But Artificial Analysis's benchmark comparison shows the two land far apart on cost per completed task: at max reasoning effort, GPT-6 Sol finishes the same benchmark task for about $1.06 versus Sonnet 5.5's $7.60, a 7.2x gap driven entirely by how many tokens each model burns to reach an answer, not by list price. Sonnet 5.5 scores higher on raw capability (56 vs 48 on Artificial Analysis's Intelligence Index at max effort), but Sol's token efficiency means it wins on cost per task except near the top of the effort range, where the two converge close to $1/task.
Receipts
- primaryhttps://artificialanalysis.ai/models/releases/comparisons/gpt-6-sol-vs-claude-sonnet-5
- alsohttps://artificialanalysis.ai/articles/claude-sonnet-5-5
- alsohttps://www.anthropic.com/claude-sonnet-5-5
- alsohttps://openai.com/index/introducing-gpt-6-sol-and-luna/
Flagged:Primary is Artificial Analysis's own release-comparison page, the independent benchmark provider both companies' own model cards cite (same data source behind this site's Leaderboard). The Intelligence Index scores (56 for Sonnet 5.5, 48 for GPT-6 Sol at max effort) and the cost-per-task figures ($7.60 vs $1.06 at max effort, a 7.2x gap) come from that page and Artificial Analysis's companion article on Sonnet 5.5. The identical $2/$10 per-million-token pricing is corroborated independently by Anthropic's own Sonnet 5.5 announcement (pricing unchanged from Sonnet 5) and OpenAI's own GPT-6 Sol announcement (pricing described elsewhere as permanent, not promotional). No claim is made about either model's current Leaderboard rank on this site — the committed snapshot predates both releases; the card points to the Leaderboard as the place to track the tradeoff going forward, not as evidence of a specific rank today.