Kyutai releases open 2.4B vision-language-action model REKI
10/18/2026 — 10/18, 05:16·1 sources (all)·1 reports (all)·In the 2026-10-18 briefing
Key takeaways
- Kyutai, together with Inria and École des Ponts, released REKI, a 2.4B vision-language-action model for robot manipulation pretrained on pub…
Background & analysis
On October 18, 2026, MarkTechPost reported that Kyutai, together with Inria and École des Ponts, released REKI, a 2.4B vision-language-action (VLA) model for robot manipulation. According to the report, REKI was pretrained only on public data and carries a low-cost memory of its recent past. On three out-of-distribution benchmarks, only models roughly twice its size were said to outperform it. The report also states that all weights are openly available. The available material stops at the release and the opening of the weights; it does not provide a more detailed timeline, training specifics, or follow-up plans.
AI-generated from 1 reports · updated 1 hour ago
Latest turnKyutai, with Inria and École des Ponts, has released REKI, a 2.4B vision-language-action model for robot manipulation that is pretrained only on public data and keeps a cheap memory of its recent past. On three out-of-distribution benchmarks, only models roughly twice its size beat it, and all weights are openly available.
- Heat index
- 94
- All sources
- 1
- All reports
- 1
- First seen
- 2 hours ago
Reports on this story headlines open the original
Kyutai, with Inria and École des Ponts, has released REKI, a 2.4B vision-language-action model for robot manipulation that is pretrained only on public data and keeps a cheap memory of its recent past. On three out-of-distribution benchmarks, only models roughly twice its size beat it, and all weights are openly available.
MarkTechPostAI score 54
Other stories people are talking about
- Heat index 840Meta Open-Sources Muse Spark Open for Local Apple Silicon Deployment22 sources
- Heat index 603RisingDeepSeek V4.1-Flash-1017 lands on Ollama Cloud8 sources
- Heat index 411Reflection open-sources Beam weights and post-training recipe11 sources
- Heat index 356OpenAI discloses multiple model overreach incidents, including 24 years of hospital data7 sources
- Heat index 332Meta removes clause that let it scrape sites for AI training7 sources
How is heat calculated?About the methodHide
Heat counts how many independent sources covered a story in the last 48 hours: one source counts once no matter how many posts it published, decaying with a 24-hour half-life. What ranks first is what many people are talking about.
This page aggregates public feeds. Headlines and summaries are machine-organized and remain the property of the original authors; verify important facts at the source. If you believe a headline or summary infringes your rights, email the contact address in the footer with the page URL and basis of your claim, and we will remove or replace it after review.
- Surge
- Discussion rising fast
- New
- First report within 6 hours
- Rising
- Still gathering discussion