← Back to all models

    Side-by-Side Comparison

    gpt-5.2-pro vs gpt-5.4

    A detailed comparison of pricing, specifications, and benchmark performance.

    OpenAIgpt-5.2-proOpenAI
    vs
    OpenAIgpt-5.4OpenAI
    Metric
    gpt-5.2-pro
    gpt-5.4
    Provider
    Provider
    OpenAI
    OpenAI
    License
    Proprietary
    Proprietary
    Release Date
    2025-08-31
    N/A
    Pricing (per 1M tokens)
    Input Price+740%
    $21.00
    $2.50
    Output Price+1020%
    $168.00
    $15.00
    Blended (1M + 1M)
    $189.00
    $17.50
    Model Details
    Context Window
    400K
    N/A
    Max Output Tokens
    128K
    N/A
    Knowledge Cutoff
    2025-08-31
    N/A
    Throughput
    N/A
    N/A
    Benchmarks
    GPQA Diamond
    87.9%
    89.4%
    Humanity's Last Exam (With Tools)
    25.4%
    27.4%
    Humanity's Last Exam (No Tools)
    35.9%
    37%
    ARC-AGI-2 Verified
    5.9%
    8.7%
    ARC-AGI-1 Verified
    70%
    76.7%
    SWE-bench Pro (Public)
    84.5%
    88.4%
    Terminal-Bench 2.0
    46.8%
    62.2%
    OSWorld Verified
    47.8%
    64.9%
    MCP Atlas
    77.8%
    86%
    tau2-bench (Telecom)
    57.7%
    71%
    MMMU-Pro (With Tools)
    94.5%
    89.3%
    MMMU-Pro (No Tools)
    86.7%
    88.6%
    GDPval
    27.4%
    28.9%
    FinanceAgent v1.1
    81.1%
    79.6%
    Investment Banking Modeling (Internal)
    84.2%
    88.3%
    OfficeQA
    43.3%
    51.1%
    Frontier Science Research
    14.9%
    15.6%
    FrontierMath (Tiers 1-3)
    8.7%
    9.4%
    OpenAI MRCR v2 8-needle (4K-8K)
    68.3%
    72%
    OpenAI MRCR v2 8-needle (32K-128K)
    51.2%
    54.4%
    OpenAI MRCR v2 8-needle (128K-512K)
    40.7%
    42.8%
    OpenAI MRCR v2 8-needle (256K-1M)
    36.1%
    39.4%
    OpenAI MRCR v2 8-needle (512K-1M)
    31.6%
    35.6%
    Graphwalks BFS (0K-128K)
    100%
    100%
    Graphwalks BFS (256K-1M)
    99.8%
    100%
    Graphwalks Parents (0K-128K)
    100%
    100%
    Graphwalks Parents (256K-1M)
    96.6%
    98%
    MMLU
    74.1%
    —
    MATH
    100%
    —
    OmniDocBench NED
    0.079
    0.045

    Verdict

    The Bottom Line

    gpt-5.4 offers both lower pricing and stronger benchmark performance across the board, making it the clear value leader in this comparison.

    Share
    ← View gpt-5.2-pro DetailsBrowse All Models →

    From the Editor

    Building an AI app?

    Skip weeks of setup. AnotherWrapper gives you 10+ production-ready AI templates with auth, payments, and APIs pre-configured.

    View AppsGet it — $249

    All Large Language Models

    OpenAI

    48 models
    computer-use-preview·GPT-3.5 Turbo·GPT-3.5-turbo·GPT-3.5-turbo-0125·GPT-3.5-turbo-1106·GPT-3.5-turbo-16k·GPT-4·GPT-4 Turbo·GPT-4-0613·GPT-4-32k·GPT-4-32k-0613·GPT-4-turbo-0125·GPT-4-turbo-1106·GPT-4-turbo-2024-04-09·gpt-4.1·gpt-4.1-mini·gpt-4.1-nano·gpt-4.5-preview·GPT-4o·GPT-4o-2024-05-13·GPT-4o-2024-08-06·gpt-4o-audio-preview·GPT-4o-mini·gpt-5·gpt-5-codex·gpt-5-mini·gpt-5-nano·gpt-5-pro·gpt-5.1·gpt-5.1-codex·gpt-5.1-codex-mini·gpt-5.2·gpt-5.3-chat-latest·gpt-5.3-codex·gpt-5.3-instant·gpt-5.4-pro·o1·o1 mini·o1 preview·o1-mini-2024-09-12·o1-preview-2024-09-12·o1-pro·o3·o3-mini·o3-pro·o4-mini

    Anthropic

    20 models
    Claude 2·Claude 2.1·Claude 3 Haiku·Claude 3 Opus·Claude 3 Sonnet·Claude 3.5 Haiku·Claude 3.5 Sonnet·Claude Haiku 3·Claude Haiku 3.5·Claude Haiku 4.5·Claude Instant 1.2·Claude Opus 3·Claude Opus 4·Claude Opus 4.5·Claude Opus 4.6·Claude Sonnet 3.5·Claude Sonnet 3.7·Claude Sonnet 4·Claude Sonnet 4.5·Claude Sonnet 4.6

    Google

    13 models
    Gemini 1.5 Flash·Gemini 1.5 Flash·Gemini 1.5 Flash 8B·Gemini 1.5 Pro·Gemini 1.5 Pro·Gemini 2.0 Flash·Gemini 2.0 Flash Lite·Gemini 2.5 Flash·Gemini 2.5 Flash Lite·Gemini 2.5 Pro·Gemini 3 Flash Preview·Gemini 3 Pro Preview·Gemini 3.1 Flash Lite Preview

    Perplexity

    5 models
    Llama 3.1 70b·R1 1776·Sonar·Sonar Deep Research·Sonar Pro

    xAI

    10 models
    Grok 2·Grok 2 Vision·Grok 3·Grok 3 Fast·Grok 3 Mini·Grok 3 Mini Fast·Grok 4·Grok 4 Fast·Grok 4.1 Fast·Grok Code Fast 1

    DeepSeek

    7 models
    DeepSeek R1·DeepSeek V2·DeepSeek V3·DeepSeek V3.1·DeepSeek V3.2·deepseek-chat·deepseek-coder

    Groq

    9 models
    DeepSeek R1 (Groq)·Llama 3.1 70b·Llama 3.1 8B (Groq)·Llama 3.3 70B (Groq)·Llama 4 Maverick (Groq)·Llama 4 Scout (Groq)·Mistral Small 2503 (Groq)·Mixtral 8x7B·Qwen QwQ 32B (Groq)

    MiniMax

    3 models
    MiniMax M2·MiniMax M2.1·MiniMax M2.5

    Qwen

    12 models
    Qwen 2.5 72B Instruct·Qwen 3 14B·Qwen 3 235B A22B·Qwen 3 30B A3B·Qwen 3 32B·Qwen 3 4B·Qwen 3 8B·Qwen 3.5 397B A17B·Qwen Max Latest·Qwen Plus Latest·Qwen QwQ Plus·Qwen Turbo Latest

    01.ai

    1 models
    Yi-Large

    AI21 Labs

    2 models
    Jamba 1.5 Large·Jamba 1.5 Mini

    Amazon

    7 models
    Amazon Nova 2 Lite·Amazon Nova 2 Omni·Amazon Nova 2 Pro·Amazon Nova Lite·Amazon Nova Micro·Amazon Nova Premier·Amazon Nova Pro

    Anyscale

    2 models
    Llama 3 70b·Mixtral 8x7B

    Cerebras

    5 models
    DeepSeek R1 (Cerebras)·DeepSeek V3 (Cerebras)·Llama 3.1 8B (Cerebras)·Llama 3.3 70B (Cerebras)·Qwen 2.5 32B (Cerebras)

    Cohere

    4 models
    Command A·Command R·Command R+·Command R7B

    DeepInfra

    6 models
    DeepSeek R1 (DeepInfra)·DeepSeek V3 (DeepInfra)·Llama 3.1 405B (DeepInfra)·Llama 3.3 70B (DeepInfra)·Mistral Small 2501 (DeepInfra)·Qwen 2.5 72B (DeepInfra)

    Fireworks

    10 models
    DeepSeek R1 (Fireworks)·DeepSeek V3.2 (Fireworks)·Kimi K2.5 (Fireworks)·Llama 3.1 405b·Llama 3.1 405B (Fireworks)·Llama 3.1 70b·Llama 4 Maverick (Fireworks)·Llama 4 Scout (Fireworks)·Mixtral 8x7B·Yi-Large

    Hyperbolic

    4 models
    DeepSeek R1 (Hyperbolic)·DeepSeek V3 (Hyperbolic)·Llama 3.1 405B (Hyperbolic)·Qwen 3 235B A22B (Hyperbolic)

    IBM Watsonx

    6 models
    Granite 3.1 2B Instruct·Granite 3.1 8B Instruct·Granite 3.2 8B Instruct·Llama 3.1 405b·Llama 3.1 70b·Mistral Large

    Meta

    6 models
    Llama 3.1 405B·Llama 3.1 70B·Llama 3.1 8B·Llama 3.3 70B·Llama 4 Maverick·Llama 4 Scout

    Mistral

    18 models
    Codestral·Magistral Medium·Magistral Small·Mistral 7B Instruct·Mistral Codestral·Mistral Devstral·Mistral Devstral 2·Mistral Devstral Small 2·Mistral Large·Mistral Large 2·Mistral Large 2411·Mistral Medium 3·Mistral Ministral 3B·Mistral Ministral 8B·Mistral Nemo·Mistral Nemo 2407·Mistral Small 3.1·Pixtral Large 2411

    Moonshot AI

    2 models
    Kimi K2 Thinking·Kimi K2.5

    Nebius AI Studio

    5 models
    DeepSeek R1 (Nebius)·DeepSeek V3 (Nebius)·Llama 3.1 405B (Nebius)·Llama 3.1 70B (Nebius)·Qwen 3 235B A22B (Nebius)

    Reka

    3 models
    Reka Core·Reka Edge·Reka Flash

    Replicate

    5 models
    DeepSeek R1 (Replicate)·Llama 3.1 405b·Llama 3.1 405B (Replicate)·Llama 3.1 70b·Mixtral 8x7B

    SambaNova Cloud

    5 models
    DeepSeek R1 (SambaNova)·DeepSeek V3 (SambaNova)·Llama 3.1 405B (SambaNova)·Llama 3.3 70B (SambaNova)·Qwen 2.5 72B (SambaNova)

    Together.AI

    10 models
    DeepSeek R1 (Together)·DeepSeek V3 (Together)·Llama 3.1 405b·Llama 3.1 405B (Together)·Llama 3.1 70b·Llama 3.3 70B (Together)·Llama 4 Maverick (Together)·Llama 4 Scout (Together)·Mixtral 8x7B·Qwen 2.5 72B (Together)

    Writer

    8 models
    Palmyra Fin 32k·Palmyra Med 32k·Palmyra X 002·Palmyra X 002 32k·Palmyra X 003·Palmyra X 004·Palmyra X 32k·Palmyra X5

    Z AI

    6 models
    GLM-4.5·GLM-4.5 Flash·GLM-4.6·GLM-4.7·GLM-5·GLM-5 Code

    From the founder

    Build
    faster with AI templates.

    AnotherWrapper gives you the foundation to build and ship fast. No more reinventing the wheel.

    Fekri — Solopreneur building AI startups
    Founder's Note

    Hi, I'm Fekri

    @fekdaoui

    Over the last 15 months, I've built around 10 different AI apps. I noticed I was wasting a lot of time on repetitive tasks like:

    • Setting up tricky APIs
    • Generating vector embeddings
    • Integrating different AI models into a flow
    • Handling user input and output
    • Authentication, paywalls, emails, ...

    So I built something to make it easy.

    Now I can build a new AI app in just a couple of hours, leveraging one of the 10+ different AI demo apps.

    10+ ready-to-use apps

    10+ AI app templates to kickstart development

    Complete codebase

    Auth, payments, APIs — all integrated

    AI-ready infrastructure

    Vector embeddings, model switching, RAG

    Production-ready

    Secure deployment, rate limiting, error handling

    Get AnotherWrapper

    One-time purchase, lifetime access

    $249

    Pay once, use forever

    View Demo AppsGet AnotherWrapper

    FAQ

    Frequently asked questions

    Have questions before getting started? Here are answers to common questions about AnotherWrapper.

    Still have questions? Email us at [email protected]