GPT-5.2 (none) AI Evaluation | LayerLens - Benchmark Results & Performance

Model

GPT-5.2 (none) by OpenAI

text+image->text

Free credits included. No credit card required.

GPT-5.2 is the latest frontier-grade model in the GPT-5 series, offering stronger agentic and long context performance compared to GPT-5.1. It uses adaptive reasoning to allocate computation dynamically, responding quickly to simple queries while spending more depth on complex tasks. Built for broad task coverage, GPT-5.2 delivers consistent gains across math, coding, science, and tool calling workloads, with more coherent long-form answers and improved tool-use reliability.

Evaluations

across 12 benchmarks

Latency

483ms

Context length

400k

tokens

Cost

$1.75 · $14

input · output per 1M tokens

Key takeaways

GPT-5.2 is designed for broad task coverage in complex, real-world problems, with a focus on adaptive reasoning across math, coding, science, and tool-calling.
It utilizes a proprietary Mixture-of-Experts (MoE) architecture, dynamically allocating computation based on query complexity, and supports text+image to text modalities.
The model demonstrates strong performance in fundamental algebraic problems (91.8% on MATH-500) and coherent long-form answers, achieving 79.07% accuracy on the GAIA benchmark for complex reasoning.
GPT-5.2 struggles significantly with geometric reasoning, nuanced visual interpretation, highly specialized scientific domains, and quantitative problems, scoring a very low 5.76% accuracy on 'Humanity's Last Exam'.
A major limitation is the significant discrepancy between its high self-reported confidence and low accuracy, particularly in specialized scientific and quantitative tasks, indicating issues with its self-assessment capabilities.

Technical specifications

Context length: 400k tokens
Max output: —
Architecture: MoE
Modalities: text+image->text
Parameters: ~1000B
Open weights: No
Licence: Proprietary
Regions: USA
Vendor: OpenAI
Provider model key: openai/gpt-5.2