Tsinghua University logo

Timer (84M)

online
thuml/timer-base-84m

84M params | 3K context | $0.00025 per forecast | Apache-2.0

Timer is THUML's 84M-parameter decoder-only Time-Series Transformer for zero-shot point forecasting. The official model card positions this checkpoint as a general-purpose univariate forecaster with context lengths up to 2880, built on the original Timer architecture rather than the later long-context extension. In the current TSFM.ai audit loop, direct upstream inference reproduces the same tendency to flatten simple repeated seasonal toy probes, so this is best treated as a checkpoint characteristic rather than a hosted-serving quirk. It is strongest once you provide at least one full 96-point patch and validate on your own seasonal regime.

Model Classification

Family

Timer

Type

time series foundation model

Pretrained time-series model exposed on TSFM.ai for zero-shot or few-shot forecasting workloads.

Training Data

260B time points from diverse domains including energy, traffic, finance, weather, and industrial sensors.

Recommended For

  • Zero-shot univariate forecasting with causal autoregressive generation
  • Long-horizon prediction across heterogeneous time-series domains

Strengths

  • TimeAttention unifies variable-length and multi-resolution inputs
  • Strong zero-shot performance from 260B-point pretraining

Limitations

  • Smaller model family with fewer checkpoint size options than Chronos or Moirai
  • Causal-only architecture limits suitability for bidirectional tasks like imputation
  • Hosted Timer serving works best once you have at least one full 96-point patch of history; very short series are a poor fit
  • The current hosted checkpoint can flatten simple repeated seasonal toy probes more than newer specialist zero-shot models

Not Ideal For

  • Histories shorter than one full 96-point patch
  • Users who need strong seasonal continuation on very small repeated-pattern probes without tuning

Capabilities

forecastingpoint-forecastingzero-shotlong-horizon

Tags

thumltimercausalzero-shotautoregressive

Specifications

Parameters
84M
Architecture
decoder-only causal transformer (Timer)
Context length
2,880
Max context
2,880
Minimum history
96
Recommended history
2,880
Input step
96 points
Required target series
1
Temperature
Ignored
Top P
Ignored
Max output
1,024
Avg latency
n/a
Uptime
n/a
Plan limits
1,000 rpm free · 1,000,000 rpm with billing
Accelerator
T4
Regions
Virginia, US
License
Apache-2.0

Pricing

Per forecast
$0.00025

Performance

Average latency
n/a
Availability
n/a
Plan limits
1,000 rpm free · 1,000,000 rpm with billing