SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2606.26421 · 2026

Otter Weather: Skillful and Computationally Efficient Medium-Range Weather Forecasting

Miles Cranmer, Richard Turner, Stratis Markou, Aliaksandra Shysheya, Jonas Scholz, Cristiana Diaconu, Payel Mukhopadhyay

arXiv · PDF · Open in the Atlas

Code that ran

6 of 8 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 2 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 4 failures of our instrument; 2 no recorded run), measured 2026-11-22; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2606.26421.json · badge: /api/ran/2606.26421.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
cambridge-mlg/otter canonical (classifier label)
classifier basis: census
6 ran · 2 no recorded run
FunctionStatusWhere it lives
compute_improvement_over_reference Ran
ran; our contract draft was wrong, not the code
cambridge-mlg/otter/otter/evaluate/plot_results.py
code served (permissive licence) · get_code("e9ff30c4f874f033")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned DataFrame. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import pandas as pd
import numpy as np

# Create a sample DataFrame with the required columns
data = {
    'name': [
        'IFS HRES vs Analysis',
        'IFS HRES vs Analysis',
        'IFS HRES vs Analysis',
        'IFS HRES vs Analysis',
        'Model A',
        'Model A',
        'Model A',
        'Model A',
        'Model B',
        'Model B',
        'Model B',
        'Model B',
    ],
    'variable': [
        'temp',
        'precip',
        'wind',
        'humidity',
        'temp',
        'precip',
        'wind',
        'humidity',
        'temp',
        'precip',
        'wind',
        'humidity',
    ],
    'lead_time': [
        24,
        24,
        24,
        24,
        24,
        24,
        24,
        24,
        24,
        24,
        24,
        24,
    ],
    'rmse': [
        2.5,
        3.2,
        1.8,
        2.1,
        2.2,
        2.8,
        1.5,
        1.9,
        2.0,
        3.0,
        1.6,
        2.0,
    ]
}

df = pd.DataFrame(data)

ARGS = {
    'df': df,
    'start_time': 24,
    'end_time': 24,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, pandas — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type DataFrame — a DataFrame output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DataFrame output the recorded ndim 2 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
dataframe_intersection_and_rmse Ran
ran; our contract draft was wrong, not the code
cambridge-mlg/otter/otter/plots/plot_score_card.py
code served (permissive licence) · get_code("c304f14bec5da801")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned DataFrame. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import pandas as pd
import numpy as np

# Create sample validation dataframes with required columns
validation_dataframe = pd.DataFrame({
    'Metric': ['RMSE', 'RMSE', 'MAE', 'MAE'],
    'Variable': ['Temperature', 'Temperature', 'Humidity', 'Humidity'],
    'Pressure Level': [500, 850, 500, 850],
    'Lead Time': [24, 24, 48, 48],
    'Value': [1.5, 2.3, 0.8, 1.2]
})

base_validation_dataframe = pd.DataFrame({
    'Metric': ['RMSE', 'RMSE', 'MAE', 'MAE'],
    'Variable': ['Temperature', 'Temperature', 'Humidity', 'Humidity'],
    'Pressure Level': [500, 850, 500, 850],
    'Lead Time': [24, 24, 48, 48],
    'Value': [1.4, 2.1, 0.75, 1.1]
})

# Rename the 'Value' column to 'Value_x' and 'Value_y' for the merge operation
# (the function expects them to be named this way after the merge)
validation_dataframe = validation_dataframe.rename(columns={'Value': 'Value_x'})
base_validation_dataframe = base_validation_dataframe.rename(columns={'Value': 'Value_y'})

# Actually, revert this - the function will do the merge and rename
validation_dataframe = pd.DataFrame({
    'Metric': ['RMSE', 'RMSE', 'MAE', 'MAE'],
    'Variable': ['Temperature', 'Temperature', 'Humidity', 'Humidity'],
    'Pressure Level': [500, 850, 500, 850],
    'Lead Time': [24, 24, 48, 48],
    'Value': [1.5, 2.3, 0.8, 1.2]
})

base_validation_dataframe = pd.DataFrame({
    'Metric': ['RMSE', 'RMSE', 'MAE', 'MAE'],
    'Variable': ['Temperature', 'Temperature', 'Humidity', 'Humidity'],
    'Pressure Level': [500, 850, 500, 850],
    'Lead Time': [24, 24, 48, 48],
    'Value': [1.4, 2.1, 0.75, 1.1]
})

ARGS = {
    'validation_dataframe': validation_dataframe,
    'base_validation_dataframe': base_validation_dataframe
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, pandas — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type DataFrame — a DataFrame output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DataFrame output the recorded ndim 2 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_norm_and_abs_max Ran
ran, honoured the contract we drafted
cambridge-mlg/otter/otter/experiment/monitoring.py
code served (permissive licence) · get_code("48c24a91814062ba")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

# Build a minimal model with parameters
class SimpleModel(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear1 = nn.Linear(4, 3)
        self.linear2 = nn.Linear(3, 2)
    
    def forward(self, x):
        return self.linear2(self.linear1(x))

# Create model instance
model = SimpleModel()

# Initialize gradients by doing a forward/backward pass
x = torch.randn(2, 4)
y = torch.randn(2, 2)
loss = model(x).sum()
loss.backward()

ARGS = {
    'model': model,
    'measurable': 'gradients'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family float — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
inv_softplus Ran
ran; our contract draft was wrong, not the code
cambridge-mlg/otter/otter/models/distributions.py
code served (permissive licence) · get_code("fe83f058a0a5c314")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned Tensor, ndim 2, dtype family float, shape (2, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# inv_softplus takes a single tensor argument x
# The function computes: x + torch.log(-torch.expm1(-x))
# This is the inverse of the softplus function
# For numerical stability, x should be positive values
# Let's create a small positive tensor

x = torch.tensor([[0.5, 1.0, 1.5, 2.0],
                   [0.3, 0.8, 1.2, 2.5]], dtype=torch.float32)

ARGS = {
    "x": x
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 2, dtype family float, shape (2, 4) — shape (2, 4) as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
metrics_to_dataframe Ran
ran; our contract draft was wrong, not the code
cambridge-mlg/otter/otter/plots/plot_score_card.py
code served (permissive licence) · get_code("891a21d8ae2c0fd9")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned DataFrame. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import pandas as pd

# Construct a valid metrics dictionary for the function
# The function expects:
# 1. 'val/temporal_resolution' key with a numeric value
# 2. Keys matching the pattern 'val/{metric}/{variable}[/{level}]/time_{step}'
#    where variable is in VAR_MAP

ARGS = {
    'metrics': {
        'val/temporal_resolution': 6,
        'val/mae/2t/time_0': 0.5,
        'val/mae/2t/time_1': 0.6,
        'val/mse/10u/time_0': 0.3,
        'val/mse/10u/time_1': 0.4,
        'val/rmse/z/500/time_0': 1.2,
        'val/rmse/z/500/time_1': 1.5,
        'val/acc/t/850/time_0': 0.95,
        'val/acc/t/850/time_1': 0.93,
        'val/bias/q/time_0': 0.01,
        'val/bias/q/time_1': 0.02,
        'val/std/sst/time_0': 0.8,
    }
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
pandas — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type DataFrame — a DataFrame output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DataFrame output the recorded ndim 2 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
process_index Ran
ran, honoured the contract we drafted
cambridge-mlg/otter/otter/plots/plot_score_card.py
code served (permissive licence) · get_code("2c21097e06193c23")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

# Setup for process_index function
ARGS = {
    'label': '2t 500'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family float — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
run_and_log_time_to_wandb Not yet run
unverified (harvested, no recorded run)
cambridge-mlg/otter/otter/experiment/monitoring.py
code served (permissive licence) · get_code("4bdb4d01915058d2")
How it ran

no invocation recorded.

run_with_timing Not yet run
unverified (harvested, no recorded run)
cambridge-mlg/otter/otter/experiment/monitoring.py
code served (permissive licence) · get_code("3c2bcafeeb48f076")
How it ran

no invocation recorded.

Also present in other papers' code

Of 8 samples lifted from this paper's linked repositories, 8 carry a function identity and 0 do not. Of the 8 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 8 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

State-of-the-art medium-range AI weather models can outperform traditional Numerical Weather Prediction (NWP) but require massive training budgets. This restricts usage for under-resourced groups and severely limits fast model iteration. Here we develop Otter Weather, a highly efficient spatiotemporal forecasting model designed to democratise high-performance weather prediction with AI. Evaluated on ERA5 reanalysis data at 1.5°resolution using standard WeatherBench protocols, the Otter family significantly advances the skill-compute Pareto frontier. The deterministic version outperforms the best NWP baseline by 9.6% at a 24-hour lead time while requiring fewer than 3.5 A100-days for training. It provides a 2× efficiency gain over lightweight AI models and a 100-fold reduction in compute compared to resource-intensive frontier architectures. We extend these efficiency gains into probabilistic forecasting by training via the Continuous Ranked Probability Score (CRPS). Scaling to a larger architecture, Otter-XL achieves a 9.7% CRPS improvement over the IFS ENS baseline. This yields an almost two-fold increase in predictive skill over comparable lightweight models at similar compute budgets. Otter-XL also outperforms frontier architectures like GenCast by over 2%, while using an order of magnitude less compute. Finally, Otter is applied out-of-the-box to a complex acoustic scattering PDE task where it outperforms a state-of-the-art foundation modelling approach, suggesting that the advances made here might apply across a range of scientific domains.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2606.26421")
get_code_for_paper("2606.26421")
have("2606.26421")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2606.26421.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2606.26421.svg)](https://syntology.ai/paper/2606.26421)

Connect an agent — have() is free.