Archive · 3 Sep 2026
Google Launches Lower-Cost Gemini 3.8 Flash AI Model
Google has released a new artificial-intelligence model called Gemini 3.8 Flash.
It is designed to help with programming, complicated questions, and tasks that require several steps.
Google says it performs as well as some bigger models made by OpenAI and Anthropic.
The model costs less for developers to use, at least during its introductory pricing period.
However, it may use more computer tokens while solving difficult problems, which can increase the total bill.
Google also released a cybersecurity version that looks for software weaknesses and suggests fixes.
The models are available through several Google services and to some Gemini app subscribers.
Other companies, including Chinese model developers, are also competing on performance and cost.
Google says Gemini 3.8 Flash improves coding, agentic tasks, and multi-step reasoning over version 3.7 Flash.
The company claims the model can rival larger systems from Anthropic and OpenAI at substantially lower token prices.
Introductory pricing runs through December 31, 2026, at $0.75 per million input tokens and $3.75 per million output tokens.
Artificial Analysis said the model was economical at its intelligence level but used about 40% more tokens than Gemini 3.7 Flash.
Google also introduced Gemini 3.8 Flash Cyber for vulnerability discovery and automated security patching.
- Who
- Google, led in the announcement by CEO Sundar Pichai, launched the models in competition with OpenAI and Anthropic.
- What
- Google introduced Gemini 3.8 Flash and Gemini 3.8 Flash Cyber, emphasizing software engineering, reasoning, vulnerability discovery, and patching.
- When
- The launch was announced on September 2, 2026; introductory pricing is available through December 31, 2026.
- Where
- The models are available through the Gemini app, Google Antigravity, AI Studio, and the Gemini API.
- Why
- Google is seeking to strengthen its position in AI coding and enterprise applications by offering improved capabilities at lower listed token prices.
Google’s Performance and Cost Claims
Independent and Competitive Caveats
Model capability
Google’s Performance and Cost Claims
Google says Gemini 3.8 Flash outperforms most larger state-of-the-art models on autonomous engineering tasks and can rival Claude Opus 5 and GPT-5.6 Sol.
Independent and Competitive Caveats
The article presents these as Google’s claims and notes that Anthropic and OpenAI are also advancing their own AI and cybersecurity models.
Overall cost
Google’s Performance and Cost Claims
Google’s listed token prices are substantially below the comparison prices given for Claude Opus 5 and GPT-5.6 Sol.
Independent and Competitive Caveats
Artificial Analysis said the model used about 40% more tokens than Gemini 3.7 Flash, while Google warned that complex tasks may raise total usage costs.
Cybersecurity use
Google’s Performance and Cost Claims
Google describes Gemini 3.8 Flash Cyber as focused on defensive vulnerability detection and automated patching, reporting strong benchmark and Chrome-security results.
Independent and Competitive Caveats
Competitors are also pursuing cybersecurity-focused systems; the article says OpenAI has claimed its upcoming Astra model surpassed critical cybersecurity benchmarks.
- Standard model
- Gemini 3.8 Flash
- Cybersecurity model
- Gemini 3.8 Flash Cyber
- Introductory input price
- $0.75 per million input tokens through December 31, 2026
- Introductory output price
- $3.75 per million output tokens through December 31, 2026
- Post-introduction pricing
- $1.50 per million input tokens and $7.50 per million output tokens
- Reported HLE-Verified score
- 54.9%, according to Google
- Reported token-use change
- Artificial Analysis found approximately 40% higher usage than Gemini 3.7 Flash