--- title: Google's third Flash in six weeks lands, Meta's Muse... description: Google shipped Gemini 3.8 Flash, its third Flash release in six weeks, with numbers strong enough that developers are openly asking why anyone would wait for a... canonical: https://daily.dev/posts/google-s-third-flash-in-six-weeks-lands-meta-s-muse-spark-undercuts-everyone-on-price-bvhu2chbj twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Google's third Flash in six weeks lands, Meta's Muse Spark undercuts everyone on price | daily.dev og:description: Google shipped Gemini 3.8 Flash, its third Flash release in six weeks, with numbers strong enough that developers are openly asking why anyone would wait for a... og:url: https://daily.dev/posts/google-s-third-flash-in-six-weeks-lands-meta-s-muse-spark-undercuts-everyone-on-price-bvhu2chbj og:image: https://api.daily.dev/og/posts/BVhU2cHBJ.png og:image:alt: Google's third Flash in six weeks lands, Meta's Muse Spark undercuts everyone on price og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Google's third Flash in six weeks lands, Meta's Muse Spark undercuts everyone on price **[Agentic Digest](https://daily.dev/sources/agents_digest)** · 4 min read · 4 upvotes · 0 comments ## Summary Google shipped Gemini 3.8 Flash, its third Flash release in six weeks, with numbers strong enough that developers are openly asking why anyone would wait for a Pro model. Meta's Muse Spark 1.3 posts near-frontier agentic scores at a fraction of Claude's price, and GLM 5.3 Flash quietly overtook DeepSeek in usage before most people even knew what it was. On the security side, a UC Berkeley paper shows hidden agent skill files can be reconstructed just by watching normal outputs, and Booz Allen found only one model completes a full autonomous cyber kill chain today. Wonderful also closed a $550M round at $5B, tripling its valuation in six months. ## Content **TLDR:** Google shipped Gemini 3.8 Flash, its third Flash release in six weeks, with numbers strong enough that developers are openly asking why anyone would wait for a Pro model. Meta's Muse Spark 1.3 posts near-frontier agentic scores at a fraction of Claude's price, and GLM 5.3 Flash quietly overtook DeepSeek in usage before most people even knew what it was. On the security side, a UC Berkeley paper shows hidden agent skill files can be reconstructed just by watching normal outputs, and Booz Allen found only one model completes a full autonomous cyber kill chain today. Wonderful also closed a $550M round at $5B, tripling its valuation in six months. --- ## Gemini 3.8 Flash ships as Google's third Flash model in six weeks The new model scores 71% on DeepSWE 1.1 (versus Opus 5's 74%) and edges out Opus 5 on HLE-Verified, 54.9% to 54.4%. Pricing holds at $0.75/$3.75 per million tokens through year-end before doubling. The real story is what this cadence implies for Google's Pro line: multiple developers are pointing out that Flash has gotten good enough that there's no obvious case left for a pricier Pro model, and Google hasn't shipped one to counter that. A gated Cyber variant is also rolling out to roughly 650 trusted partners through the Fairwind Program, claiming 2.6x better patch accuracy than larger commercial models on Chrome's internal security tests. [Read more](https://daily.dev/feed-by-ids?id=WyexJKmxW&id=8K3MHmxhX&id=lwHiWl58I) ## Muse Spark 1.3 closes in on Opus 5 at a fraction of the cost Meta's fourth Muse Spark release in five months now ranks as the top non-Anthropic model on the Artificial Analysis Intelligence Index, sitting one point behind Opus 5. It scores 98.5 on MRCR 256K-512K, beating GPT-5.6 Sol on long context, and uses 20% fewer tool calls and 25% fewer tokens than 1.2 on the same tasks. Output tokens run $4.25 per million versus $50 for Fable 5, meaning a full agentic task costs about 5.7x less. Meta still hasn't decided whether to open-weight 1.3, a decision now tangled up with EU AI Act systemic-risk exemptions. [Read more](https://daily.dev/feed-by-ids?id=sitAgWxqS&id=tX1184UvY&id=x8jQltnO4) ## Berkeley paper: you don't need an agent's hidden skill files to clone its behavior A new paper called Daydreaming shows that sending a hosted agent ordinary customer tasks and learning from the responses is enough to reconstruct 86.8% of its original behavior on new tasks, no access to prompts or SKILL.md files required. The reconstructed files often looked nothing like the originals at the file level, yet still reproduced most of the behavior. For anyone selling hosted agent skills, the practical fix is protecting the output path itself, not just locking down the prompt file. [Read more](https://daily.dev/posts/zjR7dMN4v) ## Booz Allen: only Claude Mythos completes a full autonomous cyber kill chain Booz Allen's first Cyber Weapon Index tested 18 models from US and Chinese labs, and Mythos was the only one to autonomously achieve administrator-level control with stolen credentials in every attempt, plus full domain compromise even without credentials. Grok-4.5, Muse Spark 1.1, and GLM-5.2 got full domain access but not the entire chain unassisted. A harness-equipped Claude Sonnet rivaled Mythos, and the report warns most other tested models could hit Mythos-level weaponization within six months. [Read more](https://daily.dev/posts/WZTaXk5w0) --- ## Also notable - **GLM 5.3 Flash overtakes DeepSeek in usage:** The open-weight model runs at 200+ tokens per second and costs about one-ninth the output token price of comparable models, and reportedly overtook DeepSeek in usage before most people even knew what it was. [Read more](https://daily.dev/feed-by-ids?id=6SQf3Gz7k&id=4DzVTU9e8) - **Wonderful triples valuation to $5B in six months:** The AI agents company closed a $550M Series C led by Insight Partners with Salesforce joining, up from $2B just six months ago, and grew headcount from 350 to 650 across 35+ markets. [Read more](https://daily.dev/feed-by-ids?id=AhL6cotOQ&id=uNBa3OvXL) - **DOJ backs OpenAI's fair-use defense in NYT copyright suit:** The Justice Department filed a formal statement of interest arguing that training LLMs on copyrighted text should generally qualify as fair use, partly on national-security grounds, marking Washington's first formal intervention in the AI copyright litigation wave. [Read more](https://daily.dev/posts/HkLtUDuXl) - **GitSpawn vulnerability lets malicious Git repos hijack coding agents:** Eight related vulnerabilities across seven AI coding agents, including Claude Code, Codex, and Cursor, exploit Git's core.fsmonitor config during routine commands like git status, with no need to trick the model itself; some vendors have already shipped fixes. [Read more](https://daily.dev/posts/DtZ8OT1uU) - **Claude Fable 5.1's cache pricing cuts costs 75%, but ships three breaking changes:** Cache-read pricing drops to $0.25 per million tokens with up to 45% savings on agentic workloads, but migration requires Claude Code 2.1.250+ and handling breaking changes around forced tool_choice, thinking-block portability, and message editing. [Read more](https://daily.dev/posts/oRpRjlyXT) ## Similar posts on daily.dev - [Fable 5.1, Muse Spark beats Sol, 3 new world models blow our minds](https://daily.dev/posts/fable-5-1-muse-spark-beats-sol-3-new-world-models-blow-our-minds-j1w5socny) · ThursdAI · 0 upvotes · 0 comments --- Tags: [#llm](https://daily.dev/tags/llm), [#ai-agents](https://daily.dev/tags/ai-agents), [#claude](https://daily.dev/tags/claude), [#google-gemini](https://daily.dev/tags/google-gemini), [#ai-security](https://daily.dev/tags/ai-security) [View this post on daily.dev](https://daily.dev/posts/google-s-third-flash-in-six-weeks-lands-meta-s-muse-spark-undercuts-everyone-on-price-bvhu2chbj) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"DiscussionForumPosting","mainEntityOfPage":"https://daily.dev/posts/google-s-third-flash-in-six-weeks-lands-meta-s-muse-spark-undercuts-everyone-on-price-bvhu2chbj","headline":"Google's third Flash in six weeks lands, Meta's Muse Spark undercuts everyone on price","text":"Google shipped Gemini 3.8 Flash, its third Flash release in six weeks, with numbers strong enough that developers are openly asking why anyone would wait for a Pro model. Meta's Muse Spark 1.3 posts near-frontier agentic scores at a fraction of Claude's price, and GLM 5.3 Flash quietly overtook DeepSeek in usage before most people even knew what it was. On the security side, a UC Berkeley paper shows hidden agent skill files can be reconstructed just by watching normal outputs, and Booz Allen found only one model completes a full autonomous cyber kill chain today. Wonderful also closed a $550M round at $5B, tripling its valuation in six months.","url":"https://daily.dev/posts/google-s-third-flash-in-six-weeks-lands-meta-s-muse-spark-undercuts-everyone-on-price-bvhu2chbj","datePublished":"2026-09-03T04:17:49.258Z","dateModified":"2026-09-13T20:57:29.085Z","author":{"@type":"Organization","name":"Agentic Digest","logo":"https://media.daily.dev/image/upload/s--V91DY4ls--/f_auto,q_auto/v1772617267/logos/agents_digest","url":"https://daily.dev/sources/agents_digest"},"interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":4},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"isPartOf":{"@type":"WebPage","url":"https://daily.dev/sources/agents_digest","name":"Agentic Digest"}} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Agentic Digest","item":"https://daily.dev/sources/agents_digest"},{"@type":"ListItem","position":3,"name":"Google's third Flash in six weeks lands, Meta's Muse Spark undercuts everyone on price"}]} ```