Loading video...

Video Failed to Load

Go Home

.DeepSeek-V4-Pro-0813 is live on DigitalOcean Serverless Inference and Inference Router. 🚀 1.6T params / 49B active, 1M-token context, AA Intelligence Index of 53 at max reasoning effort.

digitalocean's profile picture

DigitalOcean

222,713 subscribers

347,944 views • 27 days ago •via X (Twitter)

LIVE

Anya Rossi• Live Now

Private livecam show

5 Comments

Bob Tong's profile picture
Bob Tong24 days ago

@deepseek_ai index 53 vs the open-weight median of 27 is a genuinely wide gap, not marginal. worth checking if that holds at lower reasoning effort though, AA's number is explicitly "max reasoning effort" which usually means slower/pricier in practice than the headline suggests

RAZA | AI EXPLORER's profile picture
RAZA | AI EXPLORER25 days ago

@deepseek_ai DeepSeek V4 Pro is live on DigitalOcean. 1.6T params, 49B active, and 1M-token context.

Rao Umair's profile picture
Rao Umair27 days ago

@deepseek_ai 1.6T parameters with 49B active and a 1M-token context window is a massive setup. DeepSeek V4 Pro looks built for serious workloads.

WhiteCap Data's profile picture
WhiteCap Data26 days ago

@deepseek_ai Serverless inference cant reach the data it needs to reason about if your POS, CRM, and email systems stay isolated.

MoExperts's profile picture
MoExperts27 days ago

@deepseek_ai curious how this performs on real workloads vs the benchmark claims

Related Videos

.Kimi.ai K3 from Moonshot AI is now live on DigitalOcean Inference Engine. 1M-token context, native vision, built to run agentic tasks for hours. Supported on Inference Router and model synthesis to maximize intelligence per dollar. No setup, no model ops.
0:36

.Kimi.ai K3 from Moonshot AI is now live on DigitalOcean Inference Engine. 1M-token context, native vision, built to run agentic tasks for hours. Supported on Inference Router and model synthesis to maximize intelligence per dollar. No setup, no model ops.

DigitalOcean

665,105 views • 1 month ago

.Grok Build is on DigitalOcean Marketplace. 🆕☁️🤖 SSH in. Run grok. Build. @xAI's coding agent, pre-configured on DigitalOcean Serverless Inference.
0:38

.Grok Build is on DigitalOcean Marketplace. 🆕☁️🤖 SSH in. Run grok. Build. @xAI's coding agent, pre-configured on DigitalOcean Serverless Inference.

DigitalOcean

807,645 views • 2 months ago

Treating every inference request the same is lighting money on fire. DigitalOcean Inference Router sends each request to the right model automatically–lower cost, faster response, and no quality tradeoff. Open-source engine. Drop-in API. Live since April. ✨
0:28

Treating every inference request the same is lighting money on fire. DigitalOcean Inference Router sends each request to the right model automatically–lower cost, faster response, and no quality tradeoff. Open-source engine. Drop-in API. Live since April. ✨

DigitalOcean

594,950 views • 3 months ago

Among the fastest DeepSeek V3.2, MiniMax-M2.5, and Qwen 3.5 397B inference in the market, per Artificial Analysis benchmarks (April 2026). ⚡️🤖 Sub-1-second TTFT. 230 tokens per second. Co-designed every layer of the stack with Inferact, performance optimized vLLM, all on NVIDIA HGX B300. Live on DigitalOcean Serverless Inference now. Full breakdown in the comments. ⬇️
0:25

Among the fastest DeepSeek V3.2, MiniMax-M2.5, and Qwen 3.5 397B inference in the market, per Artificial Analysis benchmarks (April 2026). ⚡️🤖 Sub-1-second TTFT. 230 tokens per second. Co-designed every layer of the stack with Inferact, performance optimized vLLM, all on NVIDIA HGX B300. Live on DigitalOcean Serverless Inference now. Full breakdown in the comments. ⬇️

DigitalOcean

39,063 views • 4 months ago

.RadixArk brought DeepSeek V4 to production with DigitalOcean and AMD. AMD Instinct™ MI350X GPU Droplets, plus joint engineering that drove a ~10x throughput gain. When the team that builds the inference engine picks your stack, that's the signal.
0:42

.RadixArk brought DeepSeek V4 to production with DigitalOcean and AMD. AMD Instinct™ MI350X GPU Droplets, plus joint engineering that drove a ~10x throughput gain. When the team that builds the inference engine picks your stack, that's the signal.

DigitalOcean

13,381 views • 2 months ago

From POC to production inference. 🤖🌁 Join us at #DOdeploy in San Francisco where we'll discuss inference-first architectures, predictable inference economics, operational simplicity at scale, & more.
0:26

From POC to production inference. 🤖🌁 Join us at #DOdeploy in San Francisco where we'll discuss inference-first architectures, predictable inference economics, operational simplicity at scale, & more.

DigitalOcean

12,195 views • 6 months ago

DigitalOcean Inference Router is now available through OpenCode & setup is 3 steps. ☁️🆕🤝 Run/connect & it automatically routes to the right model for the job. No more paying frontier prices for a commit message.
0:33

DigitalOcean Inference Router is now available through OpenCode & setup is 3 steps. ☁️🆕🤝 Run/connect & it automatically routes to the right model for the job. No more paying frontier prices for a commit message.

DigitalOcean

20,992 views • 3 months ago

Stop overpaying for every LLM call. 💸 The DigitalOcean Inference Router resolves intent in ~200ms and beats GPT-5.1 and Claude Sonnet 4.5 on routing accuracy. Powered by Rust, WASM, and purpose-built SLMs, it adds milliseconds, not seconds, to every request. 🌊
0:40

Stop overpaying for every LLM call. 💸 The DigitalOcean Inference Router resolves intent in ~200ms and beats GPT-5.1 and Claude Sonnet 4.5 on routing accuracy. Powered by Rust, WASM, and purpose-built SLMs, it adds milliseconds, not seconds, to every request. 🌊

DigitalOcean

19,853 views • 3 months ago

Initial DeepSeek v4 and v4 Flash support added to anemll-flash-llama.cpp! 🚀 M5 MAX 128GB can run full DeepSeek-V4 with 1.6T params in original FP8/FP4 weights from SSD without requantization! One-click scripts to convert HF safetensors to dense + MoE sidecar (no GGUF needed) Inference and server examples Prefill and decoding benchmarks Branch and docs here: This is WIP and I plan to add cleanup and more testing this week. Feedback welcome!
0:21

Initial DeepSeek v4 and v4 Flash support added to anemll-flash-llama.cpp! 🚀 M5 MAX 128GB can run full DeepSeek-V4 with 1.6T params in original FP8/FP4 weights from SSD without requantization! One-click scripts to convert HF safetensors to dense + MoE sidecar (no GGUF needed) Inference and server examples Prefill and decoding benchmarks Branch and docs here: This is WIP and I plan to add cleanup and more testing this week. Feedback welcome!

Anemll

32,250 views • 4 months ago

New on DigitalOcean's Inference Engine: GLM-5.3-Flash from Z.ai. Agentic coding that rivals top-tier models, now natively multimodal for browser and computer use. 1M token context, MoE architecture, one endpoint swap to try it. Senior Developer Advocate Ryan O'Connor breaks it down, and Z.ai for Startups is there to help AI startups build faster with GLM models. 👇
1:10

New on DigitalOcean's Inference Engine: GLM-5.3-Flash from Z.ai. Agentic coding that rivals top-tier models, now natively multimodal for browser and computer use. 1M token context, MoE architecture, one endpoint swap to try it. Senior Developer Advocate Ryan O'Connor breaks it down, and Z.ai for Startups is there to help AI startups build faster with GLM models. 👇

DigitalOcean

140,583 views • 13 days ago

DigitalOcean’s model synthesis tool is now available in public preview. 🆕 A new server-side tool: run multiple models on one inference request and get a single synthesized response, no custom orchestration. Just add it to your existing inference request.
3:22

DigitalOcean’s model synthesis tool is now available in public preview. 🆕 A new server-side tool: run multiple models on one inference request and get a single synthesized response, no custom orchestration. Just add it to your existing inference request.

DigitalOcean

2,192,210 views • 1 month ago

.AMD Instinct™ MI350X GPU Droplets are live on DigitalOcean's Agentic Inference Cloud. 🆕 Lower latency. Higher throughput. ✅ Transparent pricing and production-ready scale. ✅ AI without the infrastructure headache. 💯
0:22

.AMD Instinct™ MI350X GPU Droplets are live on DigitalOcean's Agentic Inference Cloud. 🆕 Lower latency. Higher throughput. ✅ Transparent pricing and production-ready scale. ✅ AI without the infrastructure headache. 💯

DigitalOcean

25,539 views • 6 months ago

Stop prototyping, start deploying. ☁️🤖 Join us at #DOdeploy SF to master production #AI inference. Grounded in real-world costs & reliability, we'll show you how to scale with DigitalOcean. 🌁
0:25

Stop prototyping, start deploying. ☁️🤖 Join us at #DOdeploy SF to master production #AI inference. Grounded in real-world costs & reliability, we'll show you how to scale with DigitalOcean. 🌁

DigitalOcean

10,326 views • 7 months ago

LingBot-Video is out on Hugging Face MoE-based video foundation model built for embodied intelligence 30B params, only 3B active at inference Augmented with 70K hours of embodied data on top of large-scale internet video pretraining
1:02

LingBot-Video is out on Hugging Face MoE-based video foundation model built for embodied intelligence 30B params, only 3B active at inference Augmented with 70K hours of embodied data on top of large-scale internet video pretraining

AK

40,846 views • 2 months ago

I can't believe this works, but I got DeepSeek-V4-Flash (284B params) running on a Raspberry Pi 5 (8GB edition) at >1tok/s @ ~8W during full-tilt inference! It uses an untouched copy of antirez's GGUF. Took 160+ experiments over 5 days between GPT-5.5 xhigh and Opus 4.8 max.
0:32

I can't believe this works, but I got DeepSeek-V4-Flash (284B params) running on a Raspberry Pi 5 (8GB edition) at >1tok/s @ ~8W during full-tilt inference! It uses an untouched copy of antirez's GGUF. Took 160+ experiments over 5 days between GPT-5.5 xhigh and Opus 4.8 max.

Dan Woods

185,249 views • 3 months ago

.Character.AI, AMD, & #DigitalOcean are scaling the Qwen3-235B open-weight Mixture-of-Experts AI model to serve approximately 20M users. 🤖📈☁️ This required tuning a 5600/140 (ISL/OSL) workload with FP8 inference on AMD Instinct™ MI325X GPUs to maximize request throughput (QPS).
0:23

.Character.AI, AMD, & #DigitalOcean are scaling the Qwen3-235B open-weight Mixture-of-Experts AI model to serve approximately 20M users. 🤖📈☁️ This required tuning a 5600/140 (ISL/OSL) workload with FP8 inference on AMD Instinct™ MI325X GPUs to maximize request throughput (QPS).

DigitalOcean

27,726 views • 8 months ago

Berry Juicer is live! Deposits are open. Full walkthrough below. Connect a wallet, create a vault (auto sets up a uniswap v4 single sided pool), set up inference, and you're done. See how simple it is. Harvesting is automated. As your position earns fees, they're collected, converted, and credited to your inference balance on a regular cycle. No manual claiming. That balance spends across 140+ language models through Surplus Intelligence, Claude, GPT, Gemini, Grok, Llama, and more, picked per request at each model's live rate. Idle token supply becomes usable inference. Verified and open source on Base: Factory: Strategy: Wallet security by Privy Swaps and Liquidity Router : Kyber Network Built on Base Explore:
1:15

Berry Juicer is live! Deposits are open. Full walkthrough below. Connect a wallet, create a vault (auto sets up a uniswap v4 single sided pool), set up inference, and you're done. See how simple it is. Harvesting is automated. As your position earns fees, they're collected, converted, and credited to your inference balance on a regular cycle. No manual claiming. That balance spends across 140+ language models through Surplus Intelligence, Claude, GPT, Gemini, Grok, Llama, and more, picked per request at each model's live rate. Idle token supply becomes usable inference. Verified and open source on Base: Factory: Strategy: Wallet security by Privy Swaps and Liquidity Router : Kyber Network Built on Base Explore:

Berry

20,481 views • 3 months ago

Today we open-source LingBot-Video — the first MoE-based video foundation model built for embodied intelligence. 🔹30B params, only 3B active at inference. 🔹Augmented with 70K hours of embodied data on top of large-scale internet video pretraining. 🔹Already outperforming Wan2.6, Seedance 1.5 Pro, and Cosmos3 Super on RBench. 🧵👇
1:32

Today we open-source LingBot-Video — the first MoE-based video foundation model built for embodied intelligence. 🔹30B params, only 3B active at inference. 🔹Augmented with 70K hours of embodied data on top of large-scale internet video pretraining. 🔹Already outperforming Wan2.6, Seedance 1.5 Pro, and Cosmos3 Super on RBench. 🧵👇

Robbyant

1,182,795 views • 2 months ago

#Heroku is transitioning to a “sustaining engineering model”, or dev-speak for no new roadmap. 👀 Let us help you migrate your workloads to an Inference Cloud delivering trusted performance, radically simpler operations, and predictable economics—DigitalOcean. We’re ready to help you evaluate your options in a way that is realistic, structured, and built for the long term. 🌊
0:23

#Heroku is transitioning to a “sustaining engineering model”, or dev-speak for no new roadmap. 👀 Let us help you migrate your workloads to an Inference Cloud delivering trusted performance, radically simpler operations, and predictable economics—DigitalOcean. We’re ready to help you evaluate your options in a way that is realistic, structured, and built for the long term. 🌊

DigitalOcean

10,297 views • 7 months ago

I have been testing DeepSeek-V4-Pro with the Pi coding agent. I am mindblown by how well it works out of the box. A few notes: I spent a few hours building an LLM wiki with an agent powered entirely by DeepSeek-V4-Pro on Fireworks inference. This is the first time I feel like there is an open-weight model that can reason at the level of Claude and Codex. And it does this in a cost-effective way with support for 1M context length. To be clear, I am using DeepSeek-V4-Pro inside of Pi without any special configuration. It works out of the box. It's exciting that there is a model that can just be plugged into a basic harness like Pi, and it just works. I've never seen that before. Most models require lots of configuration and setup. DeepSeek's DeepSeek-V4-Pro is clearly good at agentic coding (probably the best from the open-weight models), but the model is also great on knowledge-intensive tasks where reasoning matters. The agent pulled agentic engineering best practices from different company docs (Anthropic, OpenAI, Google, Stripe, Meta, Modal, DeepSeek, Mistral, Cohere), searched and digested Reddit and HN threads, summarized arxiv papers, and surfaced trending GitHub repos. Then it distilled everything into actionable tips across categories. I love the Wiki it built. The quality is really good. Here is a snapshot of what the wiki looks like: DeepSeek-V4-Pro handled the task without breaking stride. Multi-step research queries, code generation for scaffolding, context-heavy reasoning across disparate sources. For coding specifically, this is the first open-weight model that genuinely feels like a Codex or Claude Code experience. It compares in capability and actual multi-turn agentic work. What made the loop feel so responsive was Fireworks' inference speed (the fastest in the market) and the fact that they actually validate models at the systems level before shipping. No corrupted reasoning traces. Just fast, reliable iteration. The hybrid CSA and HCA attention design cuts KV cache to just 10% and inference FLOPs by nearly 4x at 1M-token context. This is what makes the agent loop actually fast and cheap enough to run in practice. For devs who've been watching open-weight models close the gap but haven't found one that actually delivers in practice, this is the closest I've seen. Try it here:
0:31

I have been testing DeepSeek-V4-Pro with the Pi coding agent. I am mindblown by how well it works out of the box. A few notes: I spent a few hours building an LLM wiki with an agent powered entirely by DeepSeek-V4-Pro on Fireworks inference. This is the first time I feel like there is an open-weight model that can reason at the level of Claude and Codex. And it does this in a cost-effective way with support for 1M context length. To be clear, I am using DeepSeek-V4-Pro inside of Pi without any special configuration. It works out of the box. It's exciting that there is a model that can just be plugged into a basic harness like Pi, and it just works. I've never seen that before. Most models require lots of configuration and setup. DeepSeek's DeepSeek-V4-Pro is clearly good at agentic coding (probably the best from the open-weight models), but the model is also great on knowledge-intensive tasks where reasoning matters. The agent pulled agentic engineering best practices from different company docs (Anthropic, OpenAI, Google, Stripe, Meta, Modal, DeepSeek, Mistral, Cohere), searched and digested Reddit and HN threads, summarized arxiv papers, and surfaced trending GitHub repos. Then it distilled everything into actionable tips across categories. I love the Wiki it built. The quality is really good. Here is a snapshot of what the wiki looks like: DeepSeek-V4-Pro handled the task without breaking stride. Multi-step research queries, code generation for scaffolding, context-heavy reasoning across disparate sources. For coding specifically, this is the first open-weight model that genuinely feels like a Codex or Claude Code experience. It compares in capability and actual multi-turn agentic work. What made the loop feel so responsive was Fireworks' inference speed (the fastest in the market) and the fact that they actually validate models at the systems level before shipping. No corrupted reasoning traces. Just fast, reliable iteration. The hybrid CSA and HCA attention design cuts KV cache to just 10% and inference FLOPs by nearly 4x at 1M-token context. This is what makes the agent loop actually fast and cheap enough to run in practice. For devs who've been watching open-weight models close the gap but haven't found one that actually delivers in practice, this is the closest I've seen. Try it here:

elvis

60,281 views • 4 months ago