SAVRN Model Hub · Comparisons
chronos-2 vs TimeMoE-50M
Chronos-2 has 119M parameters and TimeMoE-50M has 113M parameters; both are released under Apache License 2.0; at 16-bit, chronos-2 needs about 0.3 GB (1x MI300X from $1.85 an hour) and TimeMoE-50M about 0.3 GB (1x MI300X from $1.85 an hour).
| Field | chronos-2 amazon/chronos-2 | TimeMoE-50M Maple728/TimeMoE-50M |
|---|---|---|
| Publisher | Amazon | Xiaoming Shi |
| Task | Time series forecasting | Time series forecasting |
| Modality | Time series | Time series |
| Parameters, as reported | 119M parameters | 113M parameters |
| Architecture | Chronos2Model | TimeMoeForPrediction |
| Library | chronos-forecasting | Not stated |
| Context length | Not stated | 4,096 tokens |
| Repository size | 477.9 MB | 226.8 MB |
| Artifact formats | safetensors | safetensors |
| License | apache-2.0 | apache-2.0 |
| Access | Open weights, no gate | Open weights, no gate |
| Memory at 16-bit (weights and margin) | 0.3 GB | 0.3 GB |
| Cheapest GPUs at 16-bit, per hour | 1x MI300X, $1.85 | 1x MI300X, $1.85 |
| Memory at 4-bit (weights and margin) | 0.1 GB | 0.1 GB |
| Cheapest GPUs at 4-bit, per hour | 1x MI300X, $1.85 | 1x MI300X, $1.85 |
| Revision viewed | 29ec3766d36d | 446753ee48ff |
| Downloads reported by the hub | 22.5M | 57.1k |
| Last observed | 2026-09-21 | 2026-09-18 |
An evaluation row appears only where at least two of these models report the same benchmark with the same stated configuration, metric, unit and setup. Different evaluators stay named in each cell. Values are shown as reported: no unit conversion, no ranking.
SAVRN's Notes on chronos-2
Power draw, inlet temperature, water meter reads: a facility produces time series all day, and this Amazon model forecasts them. It reads related series and covariates together and returns multi-step quantile forecasts without training on your data first. At 119M parameters it needs 0.3 GB of memory at 16-bit, and the cheapest place we price it, a single 192 GB MI300X at $1.85 per hour on-demand, is far more card than it needs, so share the card.
Apache 2.0 covers commercial use, modification and redistribution, with the duty to keep notices and state significant changes, and it carries an express patent grant from contributors. The page lists no context length, so confirm how many past points it reads per series before sizing the pipeline. Training data is autogluon/chronos_datasets and Salesforce/GiftEvalPretrain, and the Chronos-2 paper, arXiv:2510.15821, is the design reference; read it first.
Questions
Which is larger, chronos-2 or TimeMoE-50M?
chronos-2 (119M parameters) is larger than TimeMoE-50M (113M parameters), by the parameter counts their publishers report.
Which is cheaper to run, chronos-2 or TimeMoE-50M?
At 4-bit, chronos-2 fits on 1x MI300X from $1.85 an hour and TimeMoE-50M on 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.
Can I use chronos-2 commercially?
Yes. chronos-2 is released under Apache License 2.0. The Apache License 2.0 is a permissive open-source license. It permits commercial use, modification and redistribution. It requires keeping the license and copyright notices and any NOTICE file, stating significant changes, and it includes an express patent grant from contributors.
Can I use TimeMoE-50M commercially?
Yes. TimeMoE-50M is released under Apache License 2.0. The Apache License 2.0 is a permissive open-source license. It permits commercial use, modification and redistribution. It requires keeping the license and copyright notices and any NOTICE file, stating significant changes, and it includes an express patent grant from contributors.