Agentic harness that raised Claude Opus 5 from 30% to 100% on the public ARC-AGI-3 set, completing all 183 levels across 25 environments. Creator notes the public set may inflate the score.
AVO
by NVIDIA
NVIDIA agentic harness reported to lift Claude Opus 5 from 30% to 100% on the public ARC-AGI-3 set.
Updated EU status checked Last shown
AVO at a glance
- What is AVO?
- AVO is a library from NVIDIA. NVIDIA agentic harness reported to lift Claude Opus 5 from 30% to 100% on the public ARC-AGI-3 set.
- Who makes AVO?
- AVO is made by NVIDIA.
- Does AVO work in the EU?
- Unverified. No explicit evidence found. Last verified 14 Nov 2026.
- How much does AVO cost?
- Pricing unknown.
- Where was AVO shown?
- In 1 creator video by AI Search. Most recently by AI Search on 23 Aug 2026, at 29:50 in “New AI waifus, new Deepseek, realtime worlds, Happy Shrimp, tiny TTS: AI NEWS”.
Sources: creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.
AVO is an agentic harness from NVIDIA. Details are limited because no official pages could be fetched. The only information available comes from a YouTube creator's summary (AI Search), so specifics about its design, release and licensing are unconfirmed.
According to that summary, the harness raised Claude Opus 5's score on the public ARC-AGI-3 set from 30% to 100%, completing all 183 levels across 25 environments. The creator cautioned that results on the public set may overstate performance, since public environments can be tuned for or leaked into development. It is not clear whether the result carries over to the private evaluation set.
Access method, pricing and source availability could not be verified. It is categorized as a library/agent scaffold that wraps an underlying LLM rather than a standalone model.
- Agentic harness wrapped around a frontier LLM (Claude Opus 5 in the reported result)
- Reported 30% to 100% on the public ARC-AGI-3 set
- Completed 183 levels across 25 environments
- Developed by NVIDIA
- Result is on the public set only; may inflate the score
As seen on YouTube
Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.
Similar AI tools
Closest neighbours by meaning, not just by category.
DeepSeek harness
App · DeepSeek
DeepSeek's desktop agentic harness for running the DeepSeek models on coding projects, reportedly several in parallel.
Nemotron
Model · Nvidia
NVIDIA's family of open models for agentic AI: reasoning, vision, speech, retrieval and safety models.
Claude Opus 4.7
Model · Anthropic
Anthropic's Opus-tier Claude model for agentic coding and long-running knowledge work, offered via API and Claude apps.
Long Horizon Harness
Library
Backend-independent runtime that splits long-horizon agent tasks into manager, executor and auditor roles.
Claude Managed Agents
Agent · Anthropic
Anthropic's hosted service for building and running Claude-based agents with custom harnesses in the cloud.
Claude Opus 5.5
Model · Anthropic
Anthropic's Claude Opus 5.5 model for agentic coding and knowledge work, priced lower per task than Opus 5.