Pinned
DeepSeek V4 Pro 和 Flash 上线了 Go
从模型发布到接入Go，12 个小时里 DS 团队帮我们一起打磨OC的性能，跑测试，debug tool call，提PR，等我们官宣到后半夜
这种没太多流程 直接把事情做完的合作，挺难得
eh.. deepseek v4 pro is out.. 1.5T parameters.. wow
huggingface.co/deepseek-ai/De…
Frank Wang
1,681 posts
Frank Wang
@fanjiewang
opencode.ai 创始人
- Officially adding GLM-4.6 to Zen took us a bit to get everything right - now it’s: - zing zing cheap - zoom zoom fast - hosted in the US
- We're launching the `RemixSite` construct! It lets you deploy @remix_run apps to AWS with ease! A few things to highlight: - Single-Region vs Edge mode - Custom domains - Reference AWS resources in loaders/actions `process.env.API_URL`
- 24hrs since launch.. Zen doing more than a quarter of OpenRouter’s volume for Kimi K2today we're launching opencode zen zen provides the best LLMs for coding ALWAYS in the best possible configuration no more "did claude get dumber?" no more being routed to a crappy provider and we're giving it to you at cost and you can use it anywhere that enough for you?
00:00 - me + dax: let’s discuss pricing 1 hour later: ok it’s free another day, another profitable meeting
- OpenNext 2.0.0 — ISR done right ... ISR cache stored on S3 ... shared across all server functions ... revalidated async while serving stale content ... On-Demand Revalidation? Yes!
- we are working our new coding agent bugbot, code named Sum Ting Wong
- you can now use Zen anywhere - all models accessible via `/v1/chat/completions` - all models compatible with @ai-sdk/openai-compatible
- Attention rich kids: u can now reload $20, $100, or $9999 on Zen for everyone else.. stay loyal to Big Pickel
- 我们把 Kimi K2 在 OpenCode Zen 上的价格下调了 40% 编程场景 context 命中率80%多，拿着这个这数据谈了长期合同，价格平进平出转给大家 - 我不想赚钱，只想感动自己。we're making kimi k2 40% cheaper on opencode zen - lowest prices anywhere your usage has an 83% cache hit rate - coding workloads are very efficient and so we were able to negotiate a great cache token rate, do a spend commit, and pass savings to you it's only been a week btw
- Keeping ur OpenNext server warm vs running a 5min mile... which is harder... did both today 🌞 OpenNext 1.3.0 aims to fix cold start ... by giving you a `warmer-function` ... just hook it up to EventBridge scheduler ... costs $1 to warm 50 instances
