Mistral releases Timestral time-series model family
10/19/2026 — 10/20, 08:17·2 sources (all)·2 reports (all)
Key takeaways
- Mistral AI released the Timestral family of time-series foundation models, ranging from 5M to 209M parameters. According to MarkTechPost, th…
Background & analysis
On October 19, 2026, MarkTechPost reported that Mistral AI released Timestral, a family of time-series foundation models ranging from 5M to 209M parameters. According to that report, the 209M version leads on the GIFT-Eval, TIME and fev-bench benchmarks among zero-shot models and can run on a CPU.
The next day, a DEV Community post added further details: Timestral is Mistral's first time-series foundation model family, offered in four sizes from 5M to 209M parameters, with probabilistic forecasts across 21 quantiles and context up to 32,768 steps. The launch post claimed first place among pretrained zero-shot models on all three public forecasting benchmarks. The same report noted that a developer downloaded the weights from Hugging Face and ran independent tests, finding point forecasts solid but confidence intervals disappointing.
At present, Mistral's official claims and the developer's test results diverge: the company emphasizes its zero-shot ranking on the three benchmarks, while the developer questions the confidence intervals, and no further conclusion has been reached.
AI-generated from 2 reports · updated 2 hours ago
Latest turnMistral released Timestral, its first family of time-series foundation models, ranging from 5M to 209M parameters with 21-quantile probabilistic forecasts and up to 32,768-step context, claiming first place among pretrained zero-shot models on GIFT-Eval, TIME and fev-bench. A developer who ran the public weights reports that point forecasts hold up well, but the confidence intervals do not.
- Heat index
- 143
- All sources
- 2
- All reports
- 2
- First seen
- 1 day ago
Reports on this story headlines open the original
Mistral released Timestral, its first family of time-series foundation models, ranging from 5M to 209M parameters with 21-quantile probabilistic forecasts and up to 32,768-step context, claiming first place among pretrained zero-shot models on GIFT-Eval, TIME and fev-bench. A developer who ran the public weights reports that point forecasts hold up well, but the confidence intervals do not.
DEV Community · AIAI score 59
Mistral AI has released Timestral, a family of time-series foundation models ranging from 5M to 209M parameters. According to MarkTechPost, the 209M model leads GIFT-Eval, TIME and fev-bench among zero-shot models and can run on CPU.
MarkTechPostAI score 52
Other stories people are talking about
- Heat index 1065RisingMiniMax M3.1 Flash launches on Cline and gmi_cloud14 sources
- Heat index 930RisingAnthropic pledges 5% stake to public benefit fund, with top-ups as valuation grows16 sources
- Heat index 538SurgeNscale in talks with Anthropic over roughly $45B compute deal6 sources
- Heat index 473Nscale's $30B IPO valuation faces pushback as Microsoft exits UK data center9 sources
- Heat index 473RisingDHS orders federal civilian agencies to purge DeepSeek tech6 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