gpt-5.6-terra and gpt-5.6-luna match OpenAI’s price cut — Luna down 80%, Terra down 20%, effective now
Background: on July 30, OpenAI cut API prices for two models in the GPT-5.6 family, roughly three weeks after their launch. OpenAI attributes the reduction to efficiency gains made during GPT-5.6’s own development — the model rewriting and optimizing production code, plus improved token generation. APIYI has matched the new pricing; no code changes or model-name switches are needed on your side.
Price comparison (per 1M tokens, the up-to-272K tier):
gpt-5.6-luna: input $1 → $0.20, output $6 → $1.20, cached read $0.10 → $0.02 — down 80%gpt-5.6-terra: input $2.50 → $2.00, output $15 → $12.00, cached read $0.25 → $0.20 — down 20%gpt-5.6-sol: input $5 / output $30 / cached read $0.50, unchanged
gpt-5.6-luna goes from $2 / $9 to $0.40 / $1.80, and gpt-5.6-terra from $5 / $22.50 to $4 / $18. All three models remain available on the default / svip / CodexReverse groups, with both /v1/chat/completions and /v1/responses unchanged.
Full specs and tiered pricing: GPT-5.6 Terra and GPT-5.6 Luna. The cut lands hardest on high-frequency and long-context workloads — we welcome your benchmarks and feedback.
← Back to Live Updates · 📚 Monthly Archive