# Gemini 2.5 Flash ![Organization](https://img.shields.io/badge/Org-Google-4285F4) ![Context](https://img.shields.io/badge/Context-1M%20tokens-1f6feb) ![Pricing](https://img.shields.io/badge/Price-%240.15%20%2F%20%240.60%20per%201M-2da44e) ![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-Apr%202025-555) > **Quick answer:** Gemini 2.5 Flash is Google's cost-efficient tier of the Gemini 2.5 family, released April 17, 2025. It scores 61.9% on LiveCodeBench and 68.6% on Arena Hard v2, with the same 1M-token context window as Gemini 2.5 Pro at a fraction of the cost — $0.15 input / $0.60 output per 1M tokens. ## At a Glance **Where Gemini 2.5 Flash leads** - Lowest cost in the Gemini 2.5 family: $0.15/$0.60 per 1M tokens — ~8× cheaper than Pro - Full 1M-token context window — same as Gemini 2.5 Pro - 61.9% LiveCodeBench — strong coding at the cost-efficient tier - Native multimodal: text, image, audio, and video input - Thinking mode available for harder tasks - Fast latency optimised for high-throughput applications **Where it lags** - 12.08% Humanity's Last Exam vs 21.64% for Gemini 2.5 Pro - 68.6% Arena Hard v2 vs 79.0% for Pro - AetherCode 20.3% — below Pro (32.7%) on agentic web tasks - Closed API model with no open weights **Best for:** High-volume production pipelines, latency-sensitive applications, long-context batch processing, and any workflow where Pro's full capability isn't required. ## What Gemini 2.5 Flash Is Gemini 2.5 Flash is the efficiency-first tier of Google's Gemini 2.5 family. It is designed for applications where throughput, latency, and cost matter more than maximum accuracy on frontier tasks. Despite its lower price point, it shares the same 1M-token context window as Gemini 2.5 Pro — making it distinctively capable for long-context batch work at a price point comparable to o4-mini. Like Gemini 2.5 Pro, Flash supports Google's thinking mode for harder reasoning tasks, native multimodal inputs (text, image, audio, video), and is available through both Google AI Studio and Vertex AI. Its headline differentiator vs the Pro tier is speed and cost: Flash is optimised for responsiveness and throughput rather than maximum accuracy on frontier benchmarks. For teams running high-volume agentic pipelines — document processing, RAG over large corpora, automated code review at scale — Gemini 2.5 Flash provides the 1M context window of Pro at a fraction of the per-token cost. ## Specifications | Field | Value | |---|---| | Organization | Google | | Parameters | Undisclosed | | Context window | 1,000,000 tokens | | Max output | 65,536 tokens | | Architecture | Multimodal transformer (Gemini family) | | License | Proprietary (API only) | | Release date | April 17, 2025 | | Knowledge cutoff | January 2025 (estimated) | | Modality | Text + Image + Audio + Video | ## Pricing | | Input (per 1M tokens) | Output (per 1M tokens) | |---|---|---| | Google AI (≤200K input) | $0.15 | $0.60 | | Google AI (>200K input) | $0.30 | $0.90 | Thinking tokens billed as output tokens. Pricing per [Google AI pricing page](https://ai.google.dev/pricing) — verify current rates before production use. ## Context Window Gemini 2.5 Flash has a **1,000,000-token context window** — the same as Gemini 2.5 Pro, at 8× lower cost. This makes it the most cost-efficient way to access a 1M-context window among models in its class, ideal for long-document batch processing at scale. ## Public Benchmark Scores | Benchmark | Score | Source | Date | |---|---|---|---| | [LiveCodeBench](/benchmarks/livecodebench) | 61.9% | Benchgen evaluation | 2025-07 | | [Arena Hard v2](/benchmarks/arena-hard-v2) | 68.6% | Benchgen evaluation | 2025-07 | | [Humanity's Last Exam](/benchmarks/humanitys-last-exam) | 12.08% | Benchgen evaluation | 2025-07 | | [AetherCode](/benchmarks/aethercode) | 20.3% | Benchgen evaluation | 2025-07 | | [CyberGym](/benchmarks/cybergym) | 4.8% | Benchgen evaluation | 2025-07 | | [SHADE-Arena](/benchmarks/shade-arena) | 1.4 overall success | [Anthropic research post](https://www.anthropic.com/research/shade-arena-sabotage-monitoring) | 2025-06 | *Scores from Benchgen evaluations or attributed third-party sources. See the source for methodology.* ## Gemini 2.5 Flash vs Alternatives | Model | Context | LiveCodeBench | HLE | Price (in/out per 1M) | |---|---|---|---|---| | [Gemini 2.5 Pro](/models/google/gemini-2-5-pro) | 1M | 73.6% | 21.64% | $1.25 / $10 | | **Gemini 2.5 Flash** | 1M | 61.9% | 12.08% | $0.15 / $0.60 | | [o4-mini](/models/openai/o4-mini) | 200K | 74.2% | 14.28% | $1.10 / $4.40 | | [GPT-5.6 Luna](/models/openai/gpt-5-6-luna) | 1.1M | — | — | $1 / $6 | Flash is the cost leader among 1M-context models: 8× cheaper than Gemini 2.5 Pro with a modest drop in benchmark scores (~10 points on LiveCodeBench). vs o4-mini, Flash trades about 12 points on LiveCodeBench for a 5× larger context window at lower input cost. ## How Gemini 2.5 Flash Performs on Real Agent Tasks Gemini 2.5 Flash's AetherCode score of 20.3% (agentic web tasks) is below Gemini 2.5 Pro (32.7%), indicating a more significant capability gap on complex multi-step agent tasks than general benchmarks suggest. For agent workloads that require strong agentic reasoning, Pro is the better choice. For high-volume batch workflows — processing thousands of documents, running RAG at scale, or automating routine analysis tasks — Flash's price-to-performance ratio is compelling. The 1M context window enables full-document processing without chunking at a cost point that makes large-scale deployment economically viable. ## Use Gemini 2.5 Flash via API ```python import google.generativeai as genai genai.configure(api_key="YOUR_API_KEY") model = genai.GenerativeModel("gemini-2.5-flash") response = model.generate_content( "Classify and summarise the following 500 customer support tickets...", ) print(response.text) ``` ## Related Models & Benchmarks - Compare: [Gemini 2.5 Pro](/models/google/gemini-2-5-pro) · [o4-mini](/models/openai/o4-mini) · [GPT-5.6 Luna](/models/openai/gpt-5-6-luna) - Benchmarks: [LiveCodeBench](/benchmarks/livecodebench) · [Arena Hard v2](/benchmarks/arena-hard-v2) · [Humanity's Last Exam](/benchmarks/humanitys-last-exam) ## Frequently Asked Questions

What is Gemini 2.5 Flash?

Gemini 2.5 Flash is Google's fast, cost-efficient tier of the Gemini 2.5 family, released April 17, 2025. It features a 1M-token context window and native multimodal support at $0.15/$0.60 per 1M tokens — 8× cheaper than Gemini 2.5 Pro.

What is Gemini 2.5 Flash's context window?

Gemini 2.5 Flash supports a 1,000,000-token context window — the same as Gemini 2.5 Pro, at 8× lower cost. Inputs above 200K tokens are billed at a higher rate.

How much does Gemini 2.5 Flash cost?

Gemini 2.5 Flash costs $0.15 per 1M input tokens and $0.60 per 1M output tokens (for inputs up to 200K). Inputs above 200K are billed at $0.30/$0.90 per 1M.

Is Gemini 2.5 Flash open source?

No. Gemini 2.5 Flash is a proprietary, API-only model. For open-weight Google models, see the Gemma family.

What is the difference between Gemini 2.5 Flash and Gemini 2.5 Pro?

Gemini 2.5 Pro is the higher-capability tier ($1.25/$10 per 1M), scoring ~10 points higher on LiveCodeBench and ~10 points higher on Humanity's Last Exam. Flash is the cost-efficient tier ($0.15/$0.60 per 1M), optimised for throughput and latency. Both share the 1M-token context window.

--- *Specs and scores sourced from Google's official Gemini 2.5 Flash announcement (April 17, 2025) and Benchgen evaluations; third-party benchmark scores attributed inline. Pricing cited to the Google AI pricing page. Last updated 2026-07-23.*

Gemini 2.5 Flash — Training jobs

Training jobs