New AI models and hardware arrive every week. Drawing on more than 30 years of systems support experience across Asia Pacific and the UK, we pick out the releases that matter, explain them in plain English, and say what they mean for a business.
Open-weight models are AI models that you can download and run yourself: how they score, where they come from, licences and local-run notes. Updated monthly.
| DeepSeek V4.1-Flash-1017 | 32 |
| GLM-5.3 | 32 |
| MiMo-V2.6-Pro | 30 |
| Muse Spark Open | 25 |
Artificial Analysis Intelligence Index v5, 27 Oct 2026.
Open the trackerThe business trade-offs of running AI on your own premises, the security picture, and small, medium and large starting points, in plain English.
A new scoring version reset the main independent leaderboard, a high-scoring open-weight model now runs in standard software, and another has promised its weights within days. On the hardest tasks, closed models still score well above anything you can download.
Why it matters: Scores, ease of running in-house and licence questions all moved this week, so check all three before choosing a model.
Curious about AI terms? Our evolving glossary can help.
Several labs released open-weight models in October, Artificial Analysis reset its scores with Index v5, and the US ordered DeepSeek off federal civilian systems.
Read the October roundupTalk to us about where AI fits in your business.
[email protected]