Google releases Gemini 3.8 Flash, outperforming Claude Opus 5 on multiple benchmarks at lower cost
Google released Gemini 3.8 Flash, a new AI model that outperforms Anthropic's Claude Opus 5 on HLE-Verified (54.9% vs 54.4%), Terminal-Bench 2.1 (89.4% task-completion rate), and Harvey's Legal Agent Benchmark (10.0% vs 6.7%). The model also reportedly beats OpenAI's GPT-5.6 Sol and offers faster speed at lower cost, with no price increase per token.
Editorial responsibility
- No named human review is recorded for this page.
- Reports are grouped by semantic similarity and deterministic rules. Language models may assist titles, summaries, translation and cross-source analysis; the page itself is projected from evidence records.
- Current automated evidence projection
Cross-source coverage
Wire timeline
Google's Gemini 3.8 Flash model achieves 89.4% on Terminal Bench 2.1 and 73.7% on DeepSWE
A post from the account ChrisGPT announces the release of Google's Gemini 3.8 Flash model. The post highlights two benchmark scores: 89.4% on Terminal Bench 2.1, which the author notes is somewhat outdated as the benchmark is now at version 4, and 73.7% on DeepSWE verified for a flash model. The announcement is accompanied by a link to further details. The model represents a new iteration in Google's Gemini series, specifically a 'flash' variant optimized for speed or efficiency. The benchmark results position the model competitively in the AI landscape, particularly for software engineering and terminal-based tasks.
Demis Hassabis announces Gemini 3.8 Flash and 3.8 Flash Cyber AI models
Demis Hassabis, a prominent figure in AI, announced the release of Gemini 3.8 Flash, an upgrade to the AI model series, and a new variant called 3.8 Flash Cyber, which is specifically designed to advance cyber defense capabilities. The announcement, made via a post on X, highlights the rapid pace of development, noting this is another upgrade in under a month. Hassabis directed followers to a blog post for more details on the models and related cybersecurity work, emphasizing relentless progress in the field. The post includes a link to the blog for further information.
Google's Gemini 3.8 Flash beats Claude Opus 5 on HLE-Verified and Terminal-Bench benchmarks
Google has released Gemini 3.8 Flash, a new AI model that outperforms Anthropic's Claude Opus 5 on several key benchmarks. According to a post by analyst rohanpaul_ai, Gemini 3.8 Flash scored 54.9% on HLE-Verified compared to Opus 5's 54.4%. On Terminal-Bench 2.1, which tests an agent's ability to operate a terminal and complete difficult coding, security, ML, data-science, and systems tasks, Gemini 3.8 Flash achieved a 89.4% task-completion rate. On Harvey's Legal Agent Benchmark, which uses an extremely strict all-pass rule requiring every criterion to pass across complex file-based legal work, Gemini 3.8 Flash scored 10.0% versus Opus 5's 6.7%. The post notes that Google has not raised the price per token, though harder tasks may require more tokens and tool calls, increasing total cost.
Show 8 older updatesHide older updates
Google DeepMind launches Gemini 3.8 Flash on OpenRouter at 3.7 Flash's intro price
Google DeepMind has released Gemini 3.8 Flash, now available on the OpenRouter platform. The new AI model is offered at the same introductory price as its predecessor, Gemini 3.7 Flash. According to the announcement, Gemini 3.8 Flash outperforms the earlier version across all displayed benchmarks, including coding, finance, legal, video, and science domains. The post includes links to access the model directly. This release marks a competitive update in the AI model landscape, emphasizing improved performance without a price increase.
Google launches Gemini 3.8 Flash Cyber and Gemini 3.8 Flash models
Google has announced the launch of Gemini 3.8 Flash Cyber and Gemini 3.8 Flash. The Gemini 3.8 Flash Cyber model is described as the company's most capable cybersecurity model for finding and fixing vulnerabilities. It achieves a position on the Pareto Frontier on CWE-Bench for patching, indicating top-tier performance in automated vulnerability remediation. The model is being made available to trusted defenders through a new program called the Fairwind Program. The announcement was made via an X post from the account koraykv, which is associated with Google's AI efforts. The launch represents a significant step in applying large language models to cybersecurity tasks, specifically in the area of automated patching and vulnerability management.
Gemini 3.8 Flash model released for Pro and Ultra users with improved reliability
Google's GeminiApp announced the immediate availability of its newest Gemini 3.8 Flash model for Pro and Ultra users. The model is designed to deliver more reliable and comprehensive responses, providing actionable advice on everyday topics as well as handling in-depth tasks such as text analysis and complex coding. This release marks an update to the Gemini Flash line, targeting users who require enhanced performance for both routine queries and advanced technical work.
Google DeepMind launches Fairwind Program with Gemini 3.8 Flash for government cybersecurity
Google DeepMind announced the Fairwind Program, designed to help governments and trusted partners stay ahead of threats by providing access to Gemini 3.8 Flash Cyber. This initiative aims to secure vital infrastructure and protect national security. The Gemini 3.8 Flash model is rolling out now in Antigravity and via the API in Google AI Studio and Android Studio. Additionally, Google AI Pro and Ultra subscribers can access 3.8 Flash in the Gemini App and AI Mode in Google Search. The announcement highlights the company's focus on leveraging advanced AI capabilities for cybersecurity and national defense applications.
GoogleDeepMind unveils Gemini 3.8 Flash and 3.8 Flash Cyber AI models
GoogleDeepMind announced the release of two new Gemini models: 3.8 Flash and 3.8 Flash Cyber. The 3.8 Flash model is described as the company's most intelligent model yet, with significant gains over the previous 3.7 Flash version across software engineering, agentic tasks, and multi-step reasoning. The 3.8 Flash Cyber model is positioned as GoogleDeepMind's most capable cybersecurity model, featuring frontier-level vulnerability detection and automated patching capabilities. The announcement was made via an official post on X, targeting developers and organizations looking to scale AI agents and secure code.
Gemini 3.8 Flash beats GPT-5.6 Sol and Claude Opus 5 on benchmarks at lower cost
A post from the X account 'ai_for_success' claims that Google's Gemini 3.8 Flash model is outperforming OpenAI's GPT-5.6 Sol and Anthropic's Claude Opus 5 on several unspecified benchmarks. The post further asserts that Gemini 3.8 Flash achieves this performance while costing significantly less and operating at extremely high speed. The author states the model is already live and accessible in the Gemini app, providing a link to the service. This announcement, if accurate, represents a notable competitive development in the AI large language model market, positioning Google's latest offering as a cost-effective and high-performance alternative to leading models from OpenAI and Anthropic. The post does not provide specific benchmark names or numerical results.
Google releases Gemini 3.8 Flash model with Opus 5 performance at lower cost
In a post on X, user synthwavedd announced that Google has released a new AI model called Gemini 3.8 Flash. The post claims that this model provides performance comparable to Opus 5, but at a significantly lower cost and with much faster processing speed. The announcement is framed as a positive development for Google's AI offerings, with the user expressing surprise and approval. The post includes a link to further details, suggesting a formal release or benchmark results. This development is notable in the competitive AI landscape, where cost and speed are key differentiators alongside raw performance.
WSJ: Google to Release Gemini 3.8 Flash on Wednesday, Outperforms Opus 5 in Internal Tests
According to a post on X citing the Wall Street Journal, Google is set to release its Gemini 3.8 Flash model on Wednesday. The report claims that in head-to-head testing within Jetski, Google's internal coding tool, the company's engineers have preferred Gemini 3.8 Flash to Anthropic's Opus model. This suggests the new model will be on par with or superior to Opus 5, marking a significant competitive move by Google in the AI model space. The post concludes with the observation that 'Google is finally back,' indicating a positive reception to the news within the AI community.