LoginSign Up
← Back to all sparks
DigitalOcean logo

DigitalOcean

DEVOPSINFRA · APIS
Velocity5.0

Cloud provider

digitalocean.com ↗

DigitalOcean races to stock its inference cloud with every new frontier model

inference-cloudmodel-catalogagentic-workloadsopen-modelscost-efficiency
◆Current state
DigitalOcean's changelog has become a near-weekly stream of model onboarding — Nemotron 3 Ultra, Claude Opus 4.8, DeepSeek V4, Kimi K2.6, GPT-5.5 — all framed around agentic, long-running workloads. The Inference Engine, not the core cloud business, is where the visible product motion is.
◆Where it's heading
DO is positioning as a neutral, cost-competitive inference marketplace rather than betting on one model family, leaning on agentic and long-context use cases at lower cost. The cadence suggests catalog breadth and price/performance are the levers it is pulling.
◆Prediction
Expect the model-of-the-week pace to continue, with more emphasis on cost and throughput claims for agentic workloads; the open question is whether DO layers differentiated agent tooling on top or stays a pure model host.

◆Recent moves

  1. 23d ago

    Now Available: NVIDIA Nemotron 3 Ultra

    Adds NVIDIA's 550B MoE Nemotron 3 Ultra to the Inference Engine, pitched at long-running autonomous agents — the latest entry in DO's steady catalog expansion toward agentic, high-throughput workloads.

  2. 25d ago

    Now Available: Claude Opus 4.8

    Brings Anthropic's Opus 4.8 to Serverless Inference, emphasizing reliability and predictable behavior for larger autonomous tasks — consistent with DO's pattern of fast-following frontier releases.

  3. 1mo ago

    Now Available: DeepSeek-V4-Flash

    Adds DeepSeek-V4-Flash with a 1M-token context for high-throughput, multi-agent workflows, reinforcing DO's cost-efficiency angle in the model catalog.

  4. 1mo ago

    Now Available: Kimi K2.6

    Onboards Kimi K2.6, a trillion-parameter MoE model, continuing DO's push to offer frontier-scale options tuned for production efficiency.

    View source ↗
  5. 1mo ago

    Now Available: DeepSeek-V4-Pro Model

    Makes DeepSeek-V4-Pro available for planning-first, long-context agentic workflows — another incremental broadening of the inference catalog.

    View source ↗
  6. 2mo ago

    Now Available: GPT-5.5 from OpenAI

    Adds OpenAI's GPT-5.5 across DO's inference products, extending catalog coverage to the latest frontier model with the now-standard agentic framing.

    View source ↗