Ce que DigitalOcean a publié en juillet 2026

Under the Hood: Serving Kimi K3

DigitalOcean successfully launched Kimi K3 on day zero by optimizing hardware selection (NVIDIA HGX B300, AMD Instinct MI350X) and serving stack for a 2.78T-parameter model. They addressed dynamic too…

NVIDIA Nemotron

NVIDIA unveiled the Nemotron 3 family of open models with fully transparent training data, weights, and technical reports, offering hybrid Mamba-Transformer MoE architectures optimized for agentic AI.…

Kimi K3: Open Frontier Intelligence

Kimi introduced K3, a 2.8-trillion-parameter open model with native vision and a 1-million-token context window, marking the first open 3T-class model. It outperforms other tested models in long-horiz…

Outperforming Fable 5 at half the price: meet model synthesis, a new server-side tool on DigitalOcean Inference Engine

DigitalOcean introduced model synthesis, a server-side tool in its Inference Engine that orchestrates multiple open-source models in parallel to outperform single models like Fable 5 at roughly half t…

Upcoming GPU Pricing Updates

DigitalOcean will increase on-demand and 12-month reserved pricing for select NVIDIA and AMD GPUs starting August 1, 2026, citing strong demand and the need to expand high-performance compute access. …

Network Policies

Application-Specific Exclusions

Scale Faster with Managed Weaviate: Now in Public Preview on DigitalOcean

DigitalOcean introduced Managed Weaviate, a fully managed vector database service in public preview, starting at $20/month. The service handles operational tasks like backups, upgrades, and scaling wh…

DigitalOcean launches managed vector database service with full API suite

DigitalOcean introduced a new managed vector database service via its API, enabling users to provision, manage, and scale vector databases through endpoints like `/v2/vector-databases`. The service su…

DigitalOcean Evaluations: Production Model and Router Testing for the Inference Stack

DigitalOcean introduced Evaluations, a native feature in its Inference Engine, enabling teams to validate any model or inference router configuration on their own data before production. The tool supp…

Claude Claude Fable 5

Anthropic released Claude Fable 5 on July 1, 2026, restoring access after a June disruption. The model targets complex, long-running coding and knowledge work with state-of-the-art performance in codi…