Skip to content
The Chat Archive
gemini-2.5-pro-exp-03-25-grounding· Google LMArena Search Arena 24K

What are the limits for the newly announced o3 and o4-mini

  1. User

    What are the limits for the newly announced o3 and o4-mini

  2. Assistant

    Here are the limits and specifications for the newly announced o3 and o4-mini models based on the available information:

    o3 Model:

    • Context Window: 200,000 tokens[1][2]. (Note: One source mentions a 128,000 token context window for o3, but this seems less consistent with other sources[3]).
    • Maximum Output Tokens: 100,000 tokens[1][4]. (Note: One source mentions a 48,000 token output limit for o3, but this seems less consistent with other sources[3]).
    • Knowledge Cutoff: May 31, 2024[1]. (Note: One source mentions a December 2024 cutoff[3]).
    • Usage Limits (ChatGPT Plus/Team/Enterprise): 50 messages per week[5][6][7]. The weekly limit resets seven days after the first message is sent[6].
    • Usage Limits (ChatGPT Pro): "Near unlimited" access, though restrictions may apply for abusive usage or violation of terms[5][6].
    • Capabilities: Optimized for deep coding workflows and complex technical problem-solving[2]. Can agentically use tools like web search, Python analysis, visual input reasoning, and image generation[1][8][9]. Can "think with images," meaning it uses visual information in its reasoning process[9][10].

    o4-mini Model:

    • Context Window: 200,000 tokens[1][2].
    • Maximum Output Tokens: 100,000 tokens[1][4].
    • Knowledge Cutoff: May 31, 2024[1].
    • Usage Limits (ChatGPT Plus/Team/Enterprise): 150 messages per day[5][6][7].
    • Usage Limits (ChatGPT Pro): "Near unlimited" access, subject to terms of use[5][6].
    • Usage Limits (Free Users): Can be tried by selecting the "reason" or "think" button before submitting a query[5][14][10].
    • Capabilities: Optimized for fast, effective reasoning with efficient performance in coding and visual tasks[11][2]. Supports text and image inputs, generating text outputs[12][13]. Can agentically use tools like web search, Python analysis, visual input reasoning, and image generation[1][8][9]. Can "think with images"[9][10].

    Additional Notes:

    • o4-mini-high: There is also mention of an "o4-mini-high" model, which is best for coding and has a limit of 50 messages per day for ChatGPT Plus/Team/Enterprise users[5][7]. Some sources indicate this limit might be per week[7].
    • API Access: o3 and o4-mini models are available via the API for users on paid tiers, with rate limits determined by the usage tier[6][11][14]. Both models support features like function calling and structured outputs through the API[2][4].
    • Comparison: o4-mini is described as a smaller, more cost-efficient, and faster reasoning model compared to o3[9][8]. Both models show improved performance over previous models like o1 and o3-mini on various benchmarks, including coding, math, science, and visual reasoning[8][14][10].
    • Context Window vs. GPT-4o/GPT-4o mini: Both o3 and o4-mini have a larger context window (200k tokens) compared to GPT-4o and GPT-4o mini (128k tokens)[12][13][15]. However, o4-mini has a larger maximum output token limit (100k) compared to GPT-4o mini (16.4k)[12][15].

Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.