Fusion AI Consulting notes that ARC ran Astra under the Standard harness condition, where the model decides for itself which notes to carry forward between moves.
Public source
Publisher name
Public post
GPT-6 Astra dropped yesterday and the number everywhere is 98.6% on ARC-AGI-3. ARC-AGI-3 is a benchmark that drops a model into a small video game it has never seen, wit…
Company
Fusion AI Consulting
Intelligent automation for UK accountancy practices. Fusing your workflows, not replacing them.
- Industry
- IT Services and IT Consulting
- Location
- London, GB
- Company size
- 1 employee
About Fusion AI Consulting
Fusion AI Consulting partners with UK accountancy practices to free up capacity. We build automation that takes client chasing, rekeying, onboarding and MTD admin off qualified staff, so their week goes into work that bills. We also act as an external delivery partner. Internal projects sit in a queue for months. We move faster, and often remove the need for a new hire. Fixed scope, fixed price, first system live in 30 to 60 days.
See moreLatest activity
Latest activity from Fusion AI Consulting
3 signals
Discover more
Similar signals
Similar public activity from other companies.
Research & Knowledge
Silk Data
Silk Data deployed an AI document analysis system over ten months with five specialists and a platform connecting unstructured client documents to the bank's decision-making process.
Research & Knowledge
Accenture
Accenture scored 99.9% on ARC-AGI-3 with one evaluation harness, while ARC Prize reported 62.7% with the standard harness.
Research & Knowledge
ZenStatement
ZenStatement tested frontier models on 160 real accounting tasks, with 61 reconciliation tasks achieving a best overall score of 56.4%
Research & Knowledge
PrecisionAI
PrecisionAI reports that 95% of integrated AI pilots show zero P&L impact from Writer and Workplace Intelligence's 2026 research.
Research & Knowledge
Fusion Collective