NVIDIA Blackwell Leads on First Agentic AI Infrastructure Benchmark
Establishes the frame — a new agentic-infrastructure benchmark (AgentPerf) that Blackwell leads in its first published round.
AI Storyline
4 items · 2 sources · 4 days
Operational story trace
NVIDIA Blackwell
Follow in this browser to see new updates on your Live feed.
Current stateShipping
Latest change
Anthropic's Claude models are now generally available in Microsoft Foundry on Azure running on NVIDIA GB300 Blackwell Ultra GPUs, putting a frontier model on Blackwell Ultra as a managed Azure option.
Earlier contextThe story so far
Through June, NVIDIA positioned Blackwell as the substrate for agentic AI: it topped the first agentic-infrastructure benchmark (AgentPerf from Artificial Analysis) and swept MLPerf Training 6.0. Attention then moved from training to inference, where DFlash speculative decoding was reported to lift throughput up to 15x on Blackwell.
editor-curated · source-linked
Arc
Jun 12Jun 29 · now
BENCHMARK · Jun 12Blackwell tops the first agentic-AI infrastructure benchmark1 source · show source ▾
TRAINING · Jun 16Blackwell sweeps MLPerf Training 6.01 source · show source ▾
Fastest, Largest, Strongest: NVIDIA Blackwell Sweeps MLPerf Training 6.0
Extends the leadership claim from agentic inference to training with an MLPerf Training 6.0 sweep.
INFERENCE · Jun 23DFlash speculative decoding reported up to 15x faster inference on Blackwell1 source · show source ▾
Boost Inference Performance up to 15x on NVIDIA Blackwell Using DFlash Speculative Decoding - NVIDIA Developer
Pivots the thread from training to inference, citing speculative decoding (DFlash) for up to 15x serving gains on Blackwell.
NOW · Jun 29Claude goes GA on NVIDIA GB300 Blackwell Ultra in Microsoft Foundry on Azure1 source · show source ▾
Claude Meets Blackwell Ultra: Anthropic’s Models Now Run on NVIDIA GB300 in Azure
Turns benchmark leadership into product availability — Claude generally available on Azure GB300 Blackwell Ultra via Microsoft Foundry.
Day 1 Friday, Jun 12, 2026
NVIDIA Blackwell Leads on First Agentic AI Infrastructure Benchmark
Establishes the frame — a new agentic-infrastructure benchmark (AgentPerf) that Blackwell leads in its first published round.
Day 2 Tuesday, Jun 16, 2026
Fastest, Largest, Strongest: NVIDIA Blackwell Sweeps MLPerf Training 6.0
Extends the leadership claim from agentic inference to training with an MLPerf Training 6.0 sweep.
Day 3 Tuesday, Jun 23, 2026
Boost Inference Performance up to 15x on NVIDIA Blackwell Using DFlash Speculative Decoding - NVIDIA Developer
Pivots the thread from training to inference, citing speculative decoding (DFlash) for up to 15x serving gains on Blackwell.
Day 4 Monday, Jun 29, 2026
Claude Meets Blackwell Ultra: Anthropic’s Models Now Run on NVIDIA GB300 in Azure
Turns benchmark leadership into product availability — Claude generally available on Azure GB300 Blackwell Ultra via Microsoft Foundry.
What to watch — open questions
- Are the AgentPerf leadership and 15x speculative-decoding figures reproduced by independent third parties, or are they primarily NVIDIA-published?
- Will GB300 Blackwell Ultra hosting for Claude expand beyond Azure/Microsoft Foundry to other clouds?
- How do real agentic workloads price out on Blackwell Ultra versus prior-generation GPUs?
How this thread was built
editor wrote the arc · 4 beatswatcher 1 status change
Storylines are threaded mechanically from the feed: stories that share a distinctive anchor across multiple days and sources. Each item links to its original source. The evidence trace, current state, and open questions are written by the editor routine and refreshed whenever a new beat lands.