$ tail -f stew/log  ·  ← log

anthropic vs openai — the hedgehog rematch

three matchups, matched effort levels, no cherry-picking. plus what these six models actually cost.

$ tail -f stew

put the two hedgehog tests side by side: Opus 4.8 against Terra, Sonnet 5 against Luna, Fable 5 against Sol — same prompt, same effort tiers, no cherry-picking. Sol also gets a bonus extra-high frame with no Claude tier to match it against.

six hedgehogs across three matchups, drawn independently and judged blind: the side-by-side →

the numbers behind the drawings turned out to be the more interesting story. both vendors now ship a 1M-token context window with a 128K max output — Sonnet 5, Opus 4.8, and Fable 5 on one side, Luna, Terra, and Sol on the other — so the fight has moved to price and tokens-per-answer instead of headline specs. Sonnet 5 and Terra land almost on top of each other on output price. Fable 5 is the most expensive token on the board by a wide margin, but it's Anthropic's flagship — not really a fair fight against anything OpenAI currently ships.

on the one benchmark both vendors publish for every model — SWE-bench Pro — Fable 5 leads outright, and Sol trails Opus 4.8 by about 4.6 points despite output pricing $5/M higher. Luna and Terra skip the coding benchmark entirely; OpenAI is positioning them for everyday work, not head-to-head coding evals.

full price and performance breakdown, plus a cost calculator that re-sorts all six models by your own token budget: price & performance →