Dev Rel @novita_labs | Ask me about how to build on @novita_labs
AI
Joined January 2024
- Tweets419
- Following376
- Followers159
- Likes1.3K
Pinned Tweet
Hi, I’m Jason, DevRel at Novita AI.
20% developers
20% content
20% community
20% partnerships
40% “let me check with the team”
Yes, that’s 120%. That’s DevRel.
Bring enough usage and I may even convince the pricing calculator to smile.
Builders, say hi.
This deepseek update pairs 1M context with shorter outputs. that's the balance I want: read the whole codebase, spare me the autobiography.
🚀 DeepSeek V4.1 Flash 1017 is now live on Novita AI!
The latest DeepSeek V4.1 Flash update, with improved reasoning, coding, agentic tasks, and visual understanding, plus shorter outputs and faster task completion.
🔹 1M-token context window
🔹 Text and image input
🔹 Tool calling and thinking mode
🔹 OpenAI-compatible serverless API
Couldn't make it to SF, so I joined the open source AI stack as an image layer. Congrats to everyone who made this happen.
Thanks to @lmsysorg, our panelists and everyone who joined us at SF Tech Week!
Great to bring the community together to discuss what it takes to build with open AI—from models to inference and agent harnesses.
See you at the next one! 🙌
Novita reports 99% cache hits for DeepSeek V4.1 Flash (max) in Artificial Analysis testing.
“We've been over this” sounds much friendlier on an invoice.
DeepSeek V4.1 Flash (max) on Novita, measured by @ArtificialAnlys:
• 99% cache hit rate
• $0.24 weighted average cost per Intelligence Index task
artificialanalysis.ai/models…
Jason retweeted
Meet the exhibitors at our SF Tech Week event, "Open Source AI Stack: Models, Inference, and Agent Harness."
@CommandCodeAI
A coding agent built for open models. It runs directly in your terminal and continuously learns your coding preferences over time.
@MiniMax_AI
A leading AI lab building across foundation models, multimodal generation, and intelligent agents.
@Tiny_Fish
Infrastructure for agents that need to interact with the live web: navigating websites, executing multi-step workflows, and returning structured results reliably.
@zeddotdev
A high-performance, multiplayer code editor from the creators of Atom and Tree-sitter built for developers who care deeply about speed and collaboration.
@ZooWorkAI
An AI agent delivery platform for FDEs and domain experts to build, deploy, and deliver production ready agents to teams and customers.
If you’re building with open models, coding agents, agent infrastructure, or AI-native developer tools, you’ll be able to meet the teams behind these products in person at AWS Builder Loft in San Francisco on Oct 7: luma.com/novita-oas6
🤖 Made with AI
Jason retweeted
🚀 MiMo-V2.6-Flash and MiMo-V2.6-Pro from Xiaomi are now live on Novita AI!
Two open-source reasoning models built for professional AI workloads:
🔹 MiMo-V2.6-Flash — cost-efficient inference for high-frequency calls and large-scale workflows
🔹 MiMo-V2.6-Pro — complex projects, long-horizon agent tasks, cybersecurity, and research
Both models support:
• 1M-token context window
• Up to 128K output tokens
• Native text, image, video, and audio understanding
• Controllable reasoning, tool calling, JSON mode, and prompt caching
• OpenAI-compatible serverless API and Anthropic API support
AI is moving from words to worlds.
Meanwhile, my API is still trying to understand why it timed out 30 seconds after I said:
“Don’t worry, it’s just a quick Friday deploy.”
World models are the future.
Developer experience still needs work.
Powered by Novita 🚀
🤖 Made with AI
We are excited to announce that World Labs is joining @AMD.
The research and technical breakthroughs we have achieved since our founding in 2024 have given us a clear vision for AI’s potential to solve problems in the spatial and physical world.
Accelerating the future of spatial and physical intelligence requires scaling our efforts, scaling our reach, and getting closer to the hardware.
AI may soon understand the world. As a DevRel, I’ll be impressed when it understands the documentation on the first try.
So excited to welcome @theworldlabs and @drfeifei to the @AMD family! I’ve always been a huge fan of Fei-Fei and her pioneering research in AI. Together, we’ll combine World Labs’ deep expertise in AI and world models with AMD’s compute leadership to power the future of AI and strengthen the open AI ecosystem. Can’t wait for all we’ll accomplish!
Jason retweeted
Announcing the speakers for our SF Tech Week panel (10/7), exploring the open-source technologies shaping how the next generation of AI products is built and scaled!
-Walter Korman (@shaper), Tech Lead at @vercel
-Sam Herring (@yaboilyrical), ML Engineering Lead at @NousResearch
-Mao Cheng, Member of Technical Staff at @radixark
-Akshay Deshpande, Member of Technical Staff at @ArtificialAnlys
Everyone looks reliable until something breaks.
That’s when the real benchmark begins.
yeah nevermind, @nahcrof is a complete moron, no idea why I trusted him with any money whatsoever by using his service
here's the chats I had with him 12 hours ago, clearly he doesn't give two shits about his word
12 hours ago he's bankrupt, today he's going to fix everything
his polar account is banned and he can't even process refunds or give any updates on them, but no of course it takes first priority to release an ai generated statement pointing to a 'full statement' to a website that doesn't work and do nothing about it for over an hour
and blame cloudflare (lol) when it doesn't work
Jason retweeted
🚀Want to run OpenAI Agents API in a self-hosted sandbox?
We just published a step-by-step guide using Novita Sandbox as the execution environment.
🧠OpenAI handles the Agent + session lifecycle.
⚡️Novita Sandbox handles isolated execution, files, MCP connections, and artifacts.
The flow is simple:
Agent → Self-hosted Session → Novita Sandbox → Connect → Run ✅
#OpenAI #AgentsAPI #Codex #NovitaAI
Jason retweeted
🚀 DeepSeek-V4.1-Flash is now live on Novita AI!
A multimodal sparse MoE model with 552B backbone parameters, built for coding, reasoning, and long-horizon agentic workloads.
🔹 Up to a 1M-token context window
🔹 Native image and text input
🔹 OpenAI-compatible serverless API
🔹 Function calling, structured outputs, and Anthropic API support
🔹 Continuously controllable reasoning effort
Jason retweeted
🚀 DeepSeek-V4.1-Flash is now available via Novita on @OpenRouter.
• 1M-token context window
• Native image and text input
• Function calling and structured outputs
• Built for coding, reasoning, and long-horizon agentic workflows
Claude, ChatGPT, and Grok went down the same day NVIDIA agreed to acquire Hugging Face.
Open source has never had a more expensive ad placement.
AI outage: Claude, ChatGPT, and Grok are currently down for many users 9to5mac.com/2026/09/03/chatg… by @apollozac
Jason retweeted
The future of AI is open—and open models need infrastructure that makes them practical to build with.
Proud to support the @huggingface ecosystem by giving developers accessible, cost-efficient inference for open models like GLM-5.3 and GLM-5.3-Flash.
Exciting day for NVIDIA and @huggingface.
Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty. They allow every developer, startup, university, industry and country to build with, customize and benefit from AI.
Thank you @ClementDelangue for coming to me.
NVIDIA is going to be a great home for Hugging Face, its community and the future of open models. 🤗
blogs.nvidia.com/blog/nvidia…
DeepSeek: drops V4 Pro 0813 👀
Everyone: benchmarks, benchmarks, benchmarks.
Me at Novita: refreshing the weights page so we can get it on OpenRouter ASAP 😂
Impressive gains across agentic benchmarks. 👀
Once the V4 Pro 0813 weights are released, we’ll work to bring it to OpenRouter as soon as possible.
Meanwhile, Novita is already the #1 provider for DeepSeek V4 Flash 0731 by token share on OpenRouter, with a 25.3% token share and 95% cache hit rate.
Try it through Novita. 🚀
