InfronAI @InfronAI

Congrats to @jahooma and the Codebuff team on shipping free Kimi K2.7 Code! 🎉 It's running on Infron, and we're working closely with Alibaba Cloud to keep this open-source model stable and genuinely cost-effective. Great to see it live in Codebuff!

James Grugett @jahooma

3 days ago

We just launched free Kimi K2.7 Code in our free coding agent: npm i -g freebuff

42 37 568 273K 343

We just launched free Kimi K2.7 Code in our free coding agent: npm i -g freebuff

Claude Fable 5 is now available on Infron. @AnthropicAI 's Mythos-class model compressed months of engineering into days. The longer and more complex the task, the bigger its advantage. Try it now: infron.ai/models/anthrop…

Wan 2.7 is now available on Infron. @alibaba_cloud 's Thinking Mode suite and intelligent editing with first-and-last frame control. One model, four creative workflows, director-level control. Try it now: - Video edit: infron.ai/models/wan/wan… - Reference to video: infron.ai/models/wan/wan… - Image to video: infron.ai/models/wan/wan… - Text to video: infron.ai/models/wan/wan…

Claude Sonnet 5 is now available on Infron. @AnthropicAI's most agentic Sonnet closes the gap to Opus 4.8 at lower cost. Try it now: infron.ai/models/anthrop…

LongCat-2.0 is now available on Infron. @Meituan_LongCat 's agentic workhorse handles million-token contexts without choking, native tool use baked in. Long context that doesn't sacrifice reasoning. Try it now: infron.ai/models/meituan…

Excited to partner with @InfronAI 🚀 Five of our latest models are now available through the platform, including GLM-5.2, Kimi K2.7 Code, and MiniMax M2.5.

We're excited to welcome @inceptron to Infron 🎉! @inceptron is an inference compiler for the latest open models, built for price-performance. By packing more requests onto each GPU, it runs them up to 5-10x cheaper, and ships new open models within days of release. Five models are live through Inceptron, led by GLM-5.2, Kimi K2.7 Code, and MiniMax M2.5. On Infron, that's one API key. Requests route to the cheapest, fastest provider on their own, with automatic fallback. Zero model markup, and cache optimization saves more. 99.9% uptime. Try them ✨: lnkd.in/dWAR5m3n

HappyHorse 1.0 and 1.1 are now available on Infron. @HappyHorseATH's video model that ranked #1 on Artificial Analysis Video Arena in both text-to-video and image-to-video. Try them now 👇 →1.1 Text to Video: infron.ai/models/happyho… →1.1 Image to Video: infron.ai/models/happyho… →1.1 Reference to Video: infron.ai/models/happyho… →1.0 Image to Video: infron.ai/models/happyho… →1.0 Reference to Video: infron.ai/models/happyho… →1.0 Video Edit: infron.ai/models/happyho… #HappyHorse #Alibaba #VideoGeneration #TextToVideo

Kimi K2.7 Code is now available on Infron. @Kimi_Moonshot's coding model that outscores Claude Opus 4.8 on MCP tool use (81.1 vs 76.4) across real servers like GitHub, Notion, and Postgres. Try it now: 🔗 infron.ai/models/moonsho… #KimiK27 #MoonshotAI #AgenticCoding

keyboard_arrow_left Previous keyboard_arrow_right Next

GLM-5.2 is now available on Infron. @Zai_org's open-weight flagship beats GPT-5.5 on multiple long-horizon coding benchmarks at 1/6th the token cost. Try it now: 🔗 infron.ai/models/z-ai/gl… #GLM52 #ZhipuAI #OpenSource #AgenticCoding

Agnes-2.0-Flash is now available on Infron. @agnesai_sapiens 's agentic model, purpose-built for tool calling, coding, and reasoning workflows. Completely free. Try it now: 🔗 infron.ai/models/sapiens… We also offer a range of free models to get started, and a $5 minimum top-up unlocks access to everything else. #AgnesAI #SapiensAI #Freemodel

Grok Imagine Video is now available on @InfronAI @xai 's full video generation suite plus the 1.5 upgrade: better motion, better physics, better audio, all generated in the same pass. Try them now 👇 → Text to Video: infron.ai/models/x-ai/gr… → Image to Video: infron.ai/models/x-ai/gr… → Extend Video: infron.ai/models/x-ai/gr… → Reference to Video: infron.ai/models/x-ai/gr… → 1.5 Image to Video: infron.ai/models/x-ai/gr… #GrokImagine #xAI #VideoGeneration #TextToVideo

Seedance 2.0 now generates in 4K on @InfronAI . Same models. Same API. Just select 4K in the Resolution dropdown, no workflow rebuild needed. Supports all 4 modes: → Text to Video: infron.ai/models/bytedan… → Image to Video: infron.ai/models/bytedan… → Reference to Video: infron.ai/models/bytedan… → Virtual Portrait Reference to Video: infron.ai/models/bytedan… Higher resolution for landing pages, ad creative, product demos, and presenter clips. Try it → infron.ai/models

What makes your prompt cache go cold when you never even left the provider? The cached prompt is still sitting on one machine. A burst of long-context traffic sends your next request to a different machine that does not have it, so it reads nothing and pays full price. Nothing in your code changed. How do you keep it warm? We keep your calls landing on the machine that holds your cache: infron.ai/blog/sticky-ca…

Ever seen a long-context agent run fast and cheap, then slow down mid-task while the bill climbs, with nothing in your code changed? So what broke? The prompt cache. It was feeding most of the prompt back for free, then it stopped, and every step started reprocessing from scratch. We call that drop a cache cliff. Why does it happen, and how do you keep the cache warm? infron.ai/blog/cache-cli…

With the World Cup in full swing, we built an AI drone cinematography tool for stadium shots 🏟️ here's how it works. The idea: draw lines on an image to guide camera motion, then let an LLM translate those annotations into prompts. No need to describe movement in text, just draw the path. Three steps: 1. Generate a clean aerial stadium image with [Nano Banana Pro] 2. Draw a red motion path on a copy of the image in code( this is only used to communicate camera direction) 3. [Gemini 2.5 Flash Preview] reads the path and translates it into a cinematic camera prompt → fed into [Seedance 2.0 reference-to-video] ✨The key insight: the annotated image is only shown to Gemini, while Seedance receives the clean original, so no markings ever appear in the final video. ⚽️Try it on Infron → infron.ai/models/google/… infron.ai/models/bytedan… infron.ai/models/google/…

We're excited to announce a new feature: Sticky Routing. It keeps each conversation on the same healthy upstream route instead of drifting across providers, so provider-side caching stays warm. We validated it on deepseek-v4-flash: same model, same workload, cache hit rate about 14 points higher than the baseline and 22.58% lower cost overall. Method is reproducible, full numbers in the post. 👇

Andrew Zheng @AndrewHan_Zheng

4 weeks ago

x.com/i/article/2067…

0 0 1 103 0

This looks like it came straight out of a Pixar film. And the music just pulls you in. 🎧✨ The skin rendering, the glitter highlights, the curly hair. We built this with GPT Image 2 + Seedance 2.0 on @InfronAI: Step 1: Generate the DJ character with GPT Image 2 → lock the look 🎨 → infron.ai/models/openai/… Step 2: Feed that image back into GPT Image 2 → derive 6 more shots, same face, different angles Step 3: Feed all 7 into Seedance 2.0 → it scored the music, cut the edit, synced the drop 🎬 → infron.ai/models/bytedan… #Seedance #GPTImage2 #AIVideo #AITools