CodexHowSupport Us

GPT-5.5 Pro vs GPT-5.4 Pro

Green marks the cheaper side per row — informational only, never a verdict for your workload

GPT-5.5 ProGPT-5.4 Pro
Context window1,050,000 tokens1,050,000 tokens
Knowledge cutoffDec 01, 2025Aug 31, 2025
Standard input price$30.00/MTok$30.00/MTok
Standard output price$180.00/MTok$180.00/MTok
Reasoning effort levelsNot publishedNot published
Long-context repricing ruleNot publishedPublished
Tier 1 rate limit50 RPM · 50,000 TPM500 RPM · 30,000 TPM

Two premium-tier models from consecutive generations, and the differences between them aren't where the price tag would suggest to look — both cost the same at Standard, on both input and output, at every tier both publish.

Where they actually diverge

Knowledge cutoff is the clean difference: GPT-5.5 Pro's training data runs newer than GPT-5.4 Pro's. Rate limits are the less obvious one, and the more consequential: the two Pro models sit in genuinely different published limit groups, with different entry-tier RPM and TPM ceilings relative to their shared context window. Neither publishes a Fast mode option, and GPT-5.4 Pro states the long-context repricing rule explicitly while GPT-5.5 Pro's own page doesn't state it either way — recorded honestly as not published here rather than assumed to inherit its sibling's rule.

The throughput mismatch, on both

Both models carry a real gap between their large context window and their entry-tier throughput ceiling — the specific numbers differ, but the shape of the mismatch, a window far bigger than the entry-tier budget comfortably supports, shows up on both. Whichever one you pick, that's worth checking before assuming price is the only planning constraint that matters at this tier.

How to actually decide

With Standard pricing tied, this comparison comes down to recency against your account's actual rate-limit tier for each — check both models' real published RPM/TPM tables against your planned request rate before choosing on cutoff date alone.

Verified 2026-08-09 against CodexHow facts module (src/data/facts/) — see /about/#accuracy.