--- title: AI News: 6 New Models & A Nano Banana Killer | daily.dev description: A weekly roundup covers a flood of new AI models: Crea 3, an impressive new image editing model in public beta that outperforms ChatGPT image and Nano Banana... canonical: https://daily.dev/posts/ai-news-6-new-models-a-nano-banana-killer-cmtcs0dnj twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: AI News: 6 New Models & A Nano Banana Killer | daily.dev og:description: A weekly roundup covers a flood of new AI models: Crea 3, an impressive new image editing model in public beta that outperforms ChatGPT image and Nano Banana... og:url: https://daily.dev/posts/ai-news-6-new-models-a-nano-banana-killer-cmtcs0dnj og:image: https://api.daily.dev/og/posts/CMTCs0Dnj.png og:image:alt: AI News: 6 New Models & A Nano Banana Killer 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. # AI News: 6 New Models & A Nano Banana Killer **[Matt Wolfe](https://daily.dev/sources/mreflow)** · 31 min read · 1 upvotes · 0 comments ## Summary A weekly roundup covers a flood of new AI models: Crea 3, an impressive new image editing model in public beta that outperforms ChatGPT image and Nano Banana on inline edits; Black Forest Labs' open-weight Flux 3 Image Klein, runnable on consumer RTX 4070s or DGX Spark; Alibaba's Qwen Image 2.1 Turbo; and GenSpark Design, a new AI app design tool (sponsored segment). On the LLM side, Google's Gemini 4 Argon (previously announced Sept 30) now matches GPT-6 Astra and Claude Opus 5.5 on most benchmarks while costing far less per task ($2/$10 per million tokens through November). Meta's Muse Wonder scored 79% on a coding benchmark, putting it above Gemini 4 and Claude Fable 5.1, roughly tied with Claude Sonic 5.5, at pricing similar to Opus. Meta also open-sourced Muse Spark Open (816B params, 1M context, optimized for Apple M5 Ultra Mac Studios), alongside new open-weight releases from Microsoft (MAI Base 2, not yet available), Nvidia (Neotron 3.5 SuperVL, optimized for speed), and Paper Instruments (Ultramarine One Flash, cheap but competitive). OpenAI's 28-days-of-shipping campaign delivered mostly minor updates (composer predictions, mobile Codex creation, longer context windows, reliability fixes). Rapid-fire coverage includes new decision/judge models from Microsoft and Bespoke Labs, xAI's Grok Voice 3.0, ElevenLabs' instant soundtracks, Facebook's AI-tunable feed, Google's scam-text detection, a new multi-agent orchestration tool called Kylon, Microsoft's Xbox Labs AI game-mashup division, Epic Games' AI pivot, and new Apple hardware (Apple TV 4K, a home display device). ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How much does the Gemini 4 Argon API cost per million tokens? The Gemini 4 Argon API costs $2 per million input tokens and $10 per million output tokens, but that pricing is only available for a limited time through the end of November, after which the price roughly doubles. It can also be used inside Gemini on the AI Ultra or Pro paid plans, and it scores near state-of-the-art on most benchmarks while being much cheaper per task than Claude Opus 5.5. _Teams comparing model costs before committing can track pricing shifts like this on daily.dev._ ### What is Meta's Muse Wonder model and how does it score on coding benchmarks? Muse Wonder is Meta's new state-of-the-art large language model, scoring 79% on the Deep Suite coding benchmark, making it the top-scoring model there. On artificial analysis aggregated benchmarks it ranks above Gemini 4 Argon, GPT-6 Astra, and Claude Fable 5.1, roughly tied with Claude Sonic 5.5 and just two points below Claude Opus 5.5, at similar pricing of $4 per million input and $20 per million output tokens. _Developers picking a coding model can weigh benchmark results like these on daily.dev before switching._ ### Can I run the Flux 3 Image Klein model on a consumer GPU like an RTX 4070? Yes, Flux 3 Image Klein from Black Forest Labs is an open-weight model that runs on consumer hardware including an RTX 4070, doing text-to-image and image editing in roughly a second and a half slower than on a DGX Spark, which completes text-to-image in 7.3 seconds and image editing in 16.2 seconds. Weights are downloadable on Hugging Face and it can also be tested free via Hugging Face Spaces. _Anyone weighing local image-model hardware requirements can follow releases like this on daily.dev._ ## 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: [#open-source](https://daily.dev/tags/open-source), [#llm](https://daily.dev/tags/llm), [#ai-agents](https://daily.dev/tags/ai-agents), [#genai](https://daily.dev/tags/genai), [#google-gemini](https://daily.dev/tags/google-gemini) [View this post on daily.dev](https://daily.dev/posts/ai-news-6-new-models-a-nano-banana-killer-cmtcs0dnj) ```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":"TechArticle","headline":"AI News: 6 New Models & A Nano Banana Killer","url":"https://daily.dev/posts/ai-news-6-new-models-a-nano-banana-killer-cmtcs0dnj","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/ai-news-6-new-models-a-nano-banana-killer-cmtcs0dnj"},"datePublished":"2026-10-16T16:32:00.902Z","dateModified":"2026-10-16T16:57:55.191Z","description":"A weekly roundup covers a flood of new AI models: Crea 3, an impressive new image editing model in public beta that outperforms ChatGPT image and Nano Banana...","image":"https://i.ytimg.com/vi/KGcuuhqmPhI/sddefault.jpg","thumbnailUrl":"https://i.ytimg.com/vi/KGcuuhqmPhI/sddefault.jpg","isAccessibleForFree":true,"articleSection":"Matt Wolfe","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Matt Wolfe","logo":"https://media.daily.dev/image/upload/s--JWE7fI4Q--/f_auto/v1716024560/logos/mreflow","url":"https://daily.dev/sources/mreflow"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/ai-news-6-new-models-a-nano-banana-killer-cmtcs0dnj","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"open-source,llm,ai-agents,genai,google-gemini","timeRequired":"PT31M","video":{"@type":"VideoObject","name":"AI News: 6 New Models & A Nano Banana Killer","description":"A weekly roundup covers a flood of new AI models: Crea 3, an impressive new image editing model in public beta that outperforms ChatGPT image and Nano Banana...","thumbnailUrl":"https://i.ytimg.com/vi/KGcuuhqmPhI/sddefault.jpg","uploadDate":"2026-10-16T16:32:00.902Z","duration":"PT31M","url":"https://api.daily.dev/r/CMTCs0Dnj","embedUrl":"https://www.youtube.com/embed/KGcuuhqmPhI"}} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Matt Wolfe","item":"https://daily.dev/sources/mreflow"},{"@type":"ListItem","position":3,"name":"AI News: 6 New Models & A Nano Banana Killer"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/ai-news-6-new-models-a-nano-banana-killer-cmtcs0dnj#faq","mainEntity":[{"@type":"Question","name":"How much does the Gemini 4 Argon API cost per million tokens?","acceptedAnswer":{"@type":"Answer","text":"The Gemini 4 Argon API costs $2 per million input tokens and $10 per million output tokens, but that pricing is only available for a limited time through the end of November, after which the price roughly doubles. It can also be used inside Gemini on the AI Ultra or Pro paid plans, and it scores near state-of-the-art on most benchmarks while being much cheaper per task than Claude Opus 5.5. Teams comparing model costs before committing can track pricing shifts like this on daily.dev."}},{"@type":"Question","name":"What is Meta's Muse Wonder model and how does it score on coding benchmarks?","acceptedAnswer":{"@type":"Answer","text":"Muse Wonder is Meta's new state-of-the-art large language model, scoring 79% on the Deep Suite coding benchmark, making it the top-scoring model there. On artificial analysis aggregated benchmarks it ranks above Gemini 4 Argon, GPT-6 Astra, and Claude Fable 5.1, roughly tied with Claude Sonic 5.5 and just two points below Claude Opus 5.5, at similar pricing of $4 per million input and $20 per million output tokens. Developers picking a coding model can weigh benchmark results like these on daily.dev before switching."}},{"@type":"Question","name":"Can I run the Flux 3 Image Klein model on a consumer GPU like an RTX 4070?","acceptedAnswer":{"@type":"Answer","text":"Yes, Flux 3 Image Klein from Black Forest Labs is an open-weight model that runs on consumer hardware including an RTX 4070, doing text-to-image and image editing in roughly a second and a half slower than on a DGX Spark, which completes text-to-image in 7.3 seconds and image editing in 16.2 seconds. Weights are downloadable on Hugging Face and it can also be tested free via Hugging Face Spaces. Anyone weighing local image-model hardware requirements can follow releases like this on daily.dev."}}]} ```