Qwen4-Max is now on DeepInfra.
We've collaborated with @Alibaba_Qwen to bring their new flagship to our model list:
- 1M context
- text, image and video input
- tool calling, structured output, reasoning
- $1.413 / $4.24 per 1M, cached input $0.071
Try it here: deepinfra.com/Qwen/Qwen4-Max
1
1
203
Disaggregated serving, reinforcement learning, and the marriage of hardware + software.
Our CEO Nikola Borisov got into all three on the #INFERCON main stage with @nvidia and @Datadoghq
DeepInfra retweeted
We started shipping the Nvidia H100 in 2022.
A year ago it rented for $1.96/hour.
Today: $3.10.
+58%, on a chip that was supposed to be obsolete by now.
So much for GPU depreciation.
Four VFX capabilities from @ltx_io are now on Deepinfra: SDR to HDR, Alpha Gen, Reframe, and Upscale. 8-bit to 16-bit EXR, frame-matched mattes up to 4K, one master for every aspect ratio, and 4K or 8K upscales. Starting from $0.08/s. Check them out 👇
Here's what each one does:
Alpha Gen returns frame-matched grayscale mattes from existing video, up to 4K, with partial transparency so hair and soft edges hold. Compositors get a matte step without rotoscoping by hand.
SDR to HDR turns 8-bit footage into 16-bit scene-linear EXR in ACEScg, with an HLG preview for review. Archive libraries can move into an HDR-ready pipeline without going back to the original shoot.
Reframe takes one finished video and generates 1:1, 4:5, 5:4, 9:16, and 16:9 versions, filling in the missing frame content. One master serves every feed.
Upscale lifts soft or compressed footage to 4K or 8K, keeping aspect ratio, audio, and frame rate, so the master matches the delivery spec.
Try them here today:
deepinfra.com/models?q=ltx
We're excited to be at the @Sgl_project & @lmsysorg Summit in SF this week as a sponsor. Some of the best work in open-source AI infra happens in this community, and we’re looking forward to the conversations. Stop by and meet the team 👋
Sponsor booths and Full Passes for SGLang Summit 2026 are SOLD OUT! Thank you to everyone joining us. 🙌
Get your Expo Pass now to watch select main stage sessions broadcast live in the Expo area, join technical breakouts and hands-on workshops, and explore the GPU Expo. Plus, meet the teams building across the AI stack at community events.
📍 Nov 12–13 · Fort Mason, San Francisco
🎟️ Tickets & agenda: sglang.io/summit
Kimi K3.1 Pro is live on DeepInfra Day 0 🚀
Congrats to @Kimi_Moonshot on their most capable open-weight model yet.
2.8T MoE, 1M context, vision. Hosted in the US on NVIDIA Blackwell, zero data retention.
To celebrate, 5% off: $1.90 in / $9.50 out / $0.19 cached per 1M
Link in the comment ⬇️
Proud to be part of the NVIDIA Inception program, and to support the startups in it. Fellow #NVIDIAInception members get 10B free NVIDIA Nemotron tokens on DeepInfra, plus 50B more on 200+ models through DeepStart. Thank you, @NVIDIAAI
#AI
#1 US-based provider on @OpenRouter. 2.4T tokens a day. Zero data retention. No training. Thank you, builders.
NVIDIA Nemotron 3.5 Transcribe is live on DeepInfra Day 0.
Transcription, speaker attribution, translation, and audio Q&A in one open-weight model.
30B-A4B MoE. 40 ASR languages. Up to 60 minutes of audio per request. OpenAI-compatible API.
$0.002 per minute
Try it: deepinfra.com/nvidia/NVIDIA-…
2
305
Try it: deepinfra.com/nvidia/NVIDIA-…
Technical blog: deepinfra.com/blog/nvidia-ne…
82
The hard part of an AI in a group chat isn't answering. It's knowing when to stay quiet.
Boba reads the room, speaks up when it can help, and sticks with a goal for weeks.
Ampersand, then Boba, in one week. @humansand keeps shipping. Would you add an AI to your group chat? 🚀🚀🚀
Introducing Boba - the social intelligence that understands your goals and your community. Text Boba today or message it online at boba.app
Today at #NVIDIAGTC Berlin, our CEO @nikolaborisof joined @nvidia panel, “From Capacity to Customers: Building the AI Factory Business.”
The panel brought together leaders across energy, infrastructure, models, and capital. Nikola spoke for inference providers: what customers need, and how providers turn capacity into a sustainable business.
A few takeaways from the discussion:
💡 Open models give customers control. Open-weight models let teams decide how their stack runs, where their data lives, and who they depend on. That’s becoming central to sovereignty conversations.
💡 Inference at scale has to be stable and compliant. Customers moving AI into production need it to be reliable, predictable, and built to meet regulatory requirements.
💡 The AI factory is a supply chain. Financing, power, and infrastructure all have to line up before capacity turns into tokens that customers can use. From there, the business is about yield: getting the most useful output from every GPU.
Thanks to NVIDIA for hosting, and to fellow panelists @hitachienergy , @Macquarie , @MistralAI, and @nscale for a great conversation.
#AIInference #OpenSource #AIFactory



















