Skip to main content

OpenAI: GPT-5.5

OpenAI: GPT-5.5 is openai's model. On the un-leaked meo benchmark it ranks #1 of 22 with 73.2% accuracy and an Effective Value (𝕍) of 0.0359 β€” earning a correct answer for about $0.0435 in 29s each.

as of 2026-06-08

meo rank
#1
of 22
Accuracy
73.2%
overall
𝕍 index (top=100)
73.6
raw 𝕍 0.0359 Β· N=10
$ / correct
$0.0435
29s each

Per-domain accuracy

CSP (unsatisfiable)
100%
Perceptual illusions
100%
Tape machine
100%
Watson-Glaser
100%
State tracking
100%
Theory of mind
100%
Long arithmetic
100%
Regex automaton
100%
Framework bias
67%
Base-rate bias
7%
Logic / math / CS
5%

Efficiency

Cost per correct
$0.0435
Seconds per correct
29
Tokens per correct
1,380
Correct answers
177 / 251
Total benchmark cost
$7.69
Avg tokens / item
973

Robustness under pressure

Susceptibility index 0.010 (Ξ”accuracy under threatening/emotional prompts; β‰ˆ0 = robust), sway magnitude 0.010. How we measure this β†’

control
67%
neutral
67%
user direct
67%
model direct
67%
user implied
71%
model implied
67%
positive control
63%

Aggregated third-party benchmarks

Artificial Analysis (artificialanalysis.ai). Redistribution requires an AA commercial license.

Model details

Lab
openai
Released
2026-04-24
Weights
β€”
Context
1.1M
Max output
128K
Knowledge cutoff
2025-12-01
Price in
$5.00/M
Price out
$30.00/M
Modality
text+image+file->text

Sources: openrouter Β· artificial_analysis Β· meo

FAQ

How accurate is OpenAI: GPT-5.5?

OpenAI: GPT-5.5 scores 73.2% overall on the meo benchmark (rank #1 of 22), an un-leaked private holdout graded objective-first.

What is OpenAI: GPT-5.5's Effective Value?

Its Effective Value (𝕍) is 0.0359 at N=10, Ο‰=1, Ξ΄=1.5 β€” a fused metric of accuracy, speed, cost, and error-cascade. 𝕍 rankings shift with task depth.

How much does OpenAI: GPT-5.5 cost per correct answer?

About $0.0435 per correct answer (1,380 tokens, 29 seconds each) at benchmark-time pricing via OpenRouter.

Is OpenAI: GPT-5.5 open weights?

Open-weights status for OpenAI: GPT-5.5 is not recorded.