CodexHowSupport Us

GPT-5.5 — Specs, Pricing & Limits

Specs

StatusPriced
Context window1,050,000 tokens
Max output128,000 tokens
Knowledge cutoffDec 01, 2025
Reasoning effort levelsNot published
Long-context rule>272,000 input tokens: 2x input, 1.5x output (full request)

Pricing (short context)

Standard — input$5.00/MTok
Standard — cached input$0.500/MTok
Standard — cache writesNot published
Standard — output$30.00/MTok
Fast — input$12.50/MTok
Fast — output$75.00/MTok

Rate limits by usage tier

FreeNot supported
Tier 1500 RPM · 500,000 TPM · 1,500,000 batch queue
Tier 25,000 RPM · 1,000,000 TPM · 3,000,000 batch queue
Tier 35,000 RPM · 2,000,000 TPM · 100,000,000 batch queue
Tier 410,000 RPM · 4,000,000 TPM · 200,000,000 batch queue
Tier 515,000 RPM · 40,000,000 TPM · 15,000,000,000 batch queue

GPT-5.5 sits between the GPT-5.4 and GPT-5.6 generations chronologically, and its knowledge cutoff reflects that position directly — newer than GPT-5.4's, older than the current GPT-5.6 family's.

Its context window, rate-limit group and long-context repricing rule all match the current frontier generation exactly, which makes the meaningful difference between this model and its GPT-5.6 counterpart almost entirely about recency rather than capability on paper. For workloads where very recent knowledge doesn't move the outcome — well-established libraries, stable APIs, general reasoning tasks — that narrows the practical gap between choosing this model and choosing the newer one considerably.

One detail worth knowing if you're pricing a caching-heavy workload: no cache-writes price is published for this model at any service tier, which is consistent with it predating the GPT-5.6 family's introduction of a charge for writing into the cache. Caching a prefix here costs nothing extra to write, the same as on every model before the newest generation — a genuinely simpler economics than the model this site most often compares it against.

The table below is computed directly from this site's facts module: pricing across every published service tier, the long-context rule, and the rate-limit table.

Verified 2026-08-09 against https://developers.openai.com/api/docs/models/gpt-5.5.

Checked: https://developers.openai.com/api/docs/models/gpt-5.5 · https://developers.openai.com/api/docs/pricing