# GPT-5 mini ![Organization](https://img.shields.io/badge/Org-OpenAI-412991) ![License](https://img.shields.io/badge/License-Proprietary-d29922) ![Modality](https://img.shields.io/badge/Modality-Multimodal-8957e5) ![Released](https://img.shields.io/badge/Released-Aug%202025-555) > **Quick answer:** GPT-5 mini is OpenAI's cost-efficient tier of the GPT-5 family, released August 7, 2025. It scores 19.44% on Humanity's Last Exam — more than double o3's score (20.32% is comparable) and dramatically above GPT-4o mini — representing a generational leap in small-model capability. Designed to replace GPT-4o mini as the default high-throughput OpenAI model. ## At a Glance **Where GPT-5 mini leads** - 19.44% Humanity's Last Exam — frontier-class expert reasoning at the mini tier, comparable to o3 (20.32%) - GPT-5 training lineage: stronger reasoning, instruction following, and world knowledge than GPT-4o mini - Cost-efficient: designed for high-volume production pipelines - Multimodal capability - Updated knowledge cutoff (June 2025 vs GPT-4o mini's Oct 2023) **Where it lags** - Pricing not yet publicly disclosed at this page's last update — verify current rates - Lower capability ceiling than GPT-5 and GPT-5.6 Sol - No extended reasoning mode **Best for:** High-throughput pipelines, batch processing, API-integrated agents, and any workflow currently using GPT-4o mini that would benefit from dramatically improved intelligence at similar cost. ## What GPT-5 mini Is GPT-5 mini is the small-model tier of OpenAI's GPT-5 family, the successor to GPT-4o mini. Its 19.44% Humanity's Last Exam score is the most striking headline: GPT-4o mini scored approximately 0% on HLE at launch, making GPT-5 mini a step-change improvement in expert-level reasoning capability at the mini tier. The model inherits the GPT-5 training approach — which emphasises both broad world knowledge and practical reasoning — at reduced scale and cost. Like GPT-4o mini, it is designed for high-throughput applications where latency and cost-per-request matter. For teams running GPT-4o mini in production, GPT-5 mini is the natural migration target: it slots into the same cost-tier position while offering dramatically better performance on reasoning-intensive tasks, and has a significantly more recent knowledge cutoff. ## Specifications | Field | Value | |---|---| | Organization | OpenAI | | Parameters | Undisclosed | | Context window | Undisclosed | | Architecture | GPT-5 family | | License | Proprietary (API only) | | Release date | August 7, 2025 | | Knowledge cutoff | June 2025 | | Modality | Multimodal | ## Pricing Pricing for GPT-5 mini had not been publicly confirmed at the time of this page's last update. Check the [OpenAI pricing page](https://openai.com/api/pricing/) for current rates. ## Public Benchmark Scores | Benchmark | Score | Source | Date | |---|---|---|---| | [Humanity's Last Exam](/benchmarks/humanitys-last-exam) | 19.44% | Benchgen evaluation | 2025-08 | ## GPT-5 mini vs Alternatives | Model | HLE | Price (in/out per 1M) | |---|---|---| | **GPT-5 mini** | 19.44% | TBC | | [o3](/models/openai/o3) | 20.32% | $10 / $40 | | [Gemini 2.5 Pro](/models/google/gemini-2-5-pro) | 21.64% | $1.25 / $10 | | [GPT-4o mini](/models/openai/gpt-4o-mini) | ~0% | $0.15 / $0.60 | | [o4-mini](/models/openai/o4-mini) | 14.28% | $1.10 / $4.40 | GPT-5 mini's HLE score places it within 1 point of o3 (20.32%) — remarkable for a mini-tier model. This suggests that at the GPT-5 generation, the capability gap between cost-tier and premium models has narrowed substantially. ## Related Models & Benchmarks - Compare: [GPT-4o mini](/models/openai/gpt-4o-mini) · [GPT-5](/models/openai/gpt-5) · [o4-mini](/models/openai/o4-mini) · [Gemini 2.5 Flash](/models/google/gemini-2-5-flash) - Benchmarks: [Humanity's Last Exam](/benchmarks/humanitys-last-exam) ## Frequently Asked Questions

What is GPT-5 mini?

GPT-5 mini is OpenAI's cost-efficient GPT-5 tier, released August 7, 2025. It scores 19.44% on Humanity's Last Exam — dramatically above GPT-4o mini and comparable to o3 — at the mini-model price point.

Is GPT-5 mini better than GPT-4o mini?

Yes, significantly. GPT-5 mini scores 19.44% on Humanity's Last Exam vs approximately 0% for GPT-4o mini, reflecting a generational improvement in expert-level reasoning. It also has a substantially more recent knowledge cutoff (June 2025 vs October 2023).

How much does GPT-5 mini cost?

Pricing for GPT-5 mini was not publicly confirmed at this page's last update. See the OpenAI pricing page for current rates.

Is GPT-5 mini open source?

No. GPT-5 mini is a proprietary, API-only model. There are no open weights.

What is GPT-5 mini's knowledge cutoff?

GPT-5 mini's training knowledge cutoff is June 2025.

--- *Specs and scores sourced from OpenAI's official GPT-5 mini announcement (August 2025) and Benchgen evaluations. Pricing cited to the OpenAI pricing page. Last updated 2026-07-24.*

GPT-5 mini — Evaluations

Evaluations