... team spent the past week rolling Claude Sonnet 5 into a few AI agents we've been building for enterprise clients. Went in a bit skeptical honestly, most model updates lately have felt like marginal wins that don't really move the needle in production. Sonnet

... 150K 30K 80% Conversation history 200K 140K 30% Tool schemas 45K 8K 82% Total 990K 268K 73% Almost a million tokens per day, down to 268K. The quality gate ensures nothing important was lost. The Cost Breakdown Anthropic Claude Sonnet pricing: Input: $3 per million tokens Output: $15 per million tokens (unchanged by compression) Cached input: $0.30 per million tokens (90% discount) My savings come from two sources: Fewer input tokens (compression) More cache hits (prefix stabilization) Metric Before After ... ... file reads, fewer redundant tool calls, fewer wasted output tokens on re-doing work. Does Quality Actually Stay the Same? I was skeptical too. Here is what I measured over 4 weeks: Code that compiles first try: 78% (before) vs

... just say "make the tests." Ask the AI to draft a test plan, then review it. Brainstorm edge cases and iterate a few times. Once the cases are finalized, implement them using GLM (if you have), Flash, or Sonnet with "medium effort." Implementation : Whether you divide tasks by user stories, build the full stack simultaneously, or focus on the UI first, here is my model breakdown. Always ask to run test after implementation: Backend: Opus ... ... complex logic / Sonnet

... narrowed significantly over the past year. For prototyping or exploring a design, the $0.67 run is a good deal. For work requiring correctness and accuracy, Opus 4.7 remains ahead. Any experiences coding with open-weight models? [1]: MiniMax M2.7 vs. Claude Opus 4.6 Comment from fmerian(@fmerian): @yodalr  good idea, thanks! it's definitely another angle. cc @dax1 Comment from Paul Plessing(@paul_from_dentro): @amit_raj25  absolutely: Scaleway, Nebius, Inceptron for example Comment from Stan Kolotinskiy ... ... uxpin): Interesting approach - I'm usually doing it the other way around: Sonnet/Opus

... Meow world, welcome back to The Breakpoint, a weekly thread on all things dev tools on Product Hunt. The latest Recent dev-first products launched on the site @Claude by Anthropic launched Sonnet 4.6, "the most capable Sonnet

... steps. (...) To handle longer running tasks, Composer 1.5 has the ability to self-summarize. This allows the model to continue exploring for a solution even when it runs out of available context. In a previous thread, Claude's (Opus, Sonnet) and Google's (Gemini) coding models were leading the way. [2] Curious if this new release would attract more developers. Any experiences with Composer 1.5? [1]: Introducing Composer 1.5 [2]: What's the best AI model for coding? Comment from ... ... Codex pricing (per 1M tokens): Input + Cache Write: $1.75 Output: $14 Cache Read: $0.175 Comment from Vladimir(@kelindev): Hi, devs! There's no Composer 2 thread so I'd post here. I like Composer 2 more than Sonnet

... better reduce, control, and make the ecosystem a bit more frugal... at least I hope so. Comment from LadyF AI(@ladyf_ai): Opus 4.5. The only downside is it consumes very quickly. So my second choice is sonnet 4.5 Comment from Kashyap Rathod(@kashyaprathod): Claude Sonnet

... amongst this community of builders. I ll kick off with an experience that had us scratching our heads for months last year Our product, @Tonkotsu , runs a bunch of coding agents in parallel. They re built on top of Sonnet 4.5 and do coding tasks by repeatedly calling tools to read and write code. We review task failures daily and started noticing something strange: some sessions would devolve into an infinite loop of Sonnet

... generating a playable "Minecraft" replica demo cost less than 0.2 RMB using this model. Performance Benchmark: In major programming evaluation sets, Doubao-Seed-Code s overall performance is reported to be only second to the current top model, Claude Sonnet 4.5 , outperforming several other prominent domestic models. Massive Context: It features a native 256K context window , which is larger than Claude Sonnet

... just launched AllPub.co and built pretty much the entire thing using Vibecoding with Claude Sonnet 3.5 . I'm a software architect with just some basic Python knowledge, so this was my first full web app build. Wanted to share my experience! What I Built AllPub.co is a tool that lets you write once in your favourite platform or Notion and automatically publish to Dev.to and Hashnode. I'm currently exploring more integrations like Medium, LinkedIn, and others based on what ... ... people actually want. No more copy-pasting and reformatting the same article three times. Why Claude Sonnet

... Anthropic just released two new models for Claude 4: Opus 4 and Sonnet 4 . Let's find out how good they are. What are you using it for? What s impressed you? What s broken? How does it compare to GPT-4, Gemini, and others? Drop your thoughts, screenshots, prompts, use cases below. Let s trade notes while it s hot Comment from Robynne M.(@robynne_o1): @steveb   @parth_ahir  This ^^^ Comment from Charlotte Taylor(@deleted-7463904): what ... ... Willame(@jean_willame): Sadly Claude Opus 4 is on the MAX plan on Cursor :( But I can already see more relevant content from my claude desktop project either. Comment from Parth Ahir(@parth_ahir): Claude 4 s Opus and Sonnet

... Polite AI is excited to announce the integration of Anthropic s latest AI models, Claude Opus 4 and Claude Sonnet 4. These models are designed to enhance coding, reasoning, and agentic capabilities. Key Features: Claude Opus 4: Anthropic s most advanced model, capable of sustained performance on complex, long-running tasks and agent workflows. Claude Sonnet

... really bad coding assistant. I used it when the task was straight forward and simple (because it was free), but for anything that was complex or might touch more than one file, I fell back on Sonnet. I found Sonnet

... collaborative and productive environment. It offers participants the freedom to move around the meeting space (designed as various maps, some inspired by games and some based off of actual office blueprints). Nonilion is a distraction free google s meet alternative, with perplexity ai, multi participant screen share, interactive multi participant games, file sharing, custom characters and more. Nonilion is free to use for all the users and also offers a pro plan for meetings that required custom room names ... ... writers, video editors and front-end developers to focus and work in collaboration. Two or more groups talking over each other or sharing screens at the same time resulted in other groups to sit and just watch. As an alternative

... Codebase Bolt works well initially but can struggle as your codebase grows beyond roughly 100-200 prompts. Keeping your components short (under 300 lines each) helps manage complexity and Bolt's understanding of your code. Bolt uses the Claude Sonnet 3.7 model, which tends to be overly creative, suggesting unnecessarily complex solutions. Explicitly instruct Bolt, "I value simplicity and prefer straightforward implementations," to guide it towards simpler, practical solutions. Giving Bolt ambiguous prompts with multiple possible implementations often results ... ... fairly loaded topic. Analytics & Building Your Own Admin Tools Easily integrate Google Analytics by simply inputting your measurement ID into Bolt for basic user insights. Remember to ask it to track key events across your application. Alternatively

... Just a thought on Cursor and the fact that Anthropic rate limiting basically doesn't let me do anything right now. I've been developing with Cursor and Claude-3.5-Sonnet (recently 3.7-thinking) for last few months now. I've gotten to a point where I realized 1 hour of no-Cursor coding can be done in 10 minutes or less using Cursor agent. It's absolutely amazing, let me just say that. But because of the issues with ... ... Anthropic's endpoint, I was forced to try out gpt-4o, and other models. But nothing beats Claude-3.5(7)-Sonnet

... editor journey has been: 1. Tried out Github copilot in VS Code. Loved it. 2. Then tried out @Cursor after @bernatfortet kept talking about it. Immediately switch to Cursor. 3. Then tried out @Windsurf after @lagap discovered it. Immediately fell in love and switched. Anecdotally, everyone is talking about Cursor. And no one is really talking about Windsurf. On Product Hunt, Cursor has 845 Shoutouts. Windsurf has two? In my experience, Windsurf is the leader at the moment with ... ... Because it's stable, they listen to feedback, and have unlimited premium prompts on a subscription. Comment from Jo o Ferreira(@linklogician): Initially tried @Cursor , but ended up building most of my latest tool with @Windsurf . Started using Claude 3.5 Sonnet