EcoLogitsGenAIExternalAPIJob
One inference call against an EcoLogitsGenAIExternalAPI, sized by the average output token count. Per-request energy and embodied GWP are computed from the EcoLogits impact DAG.
Params
name
A human readable description of the object.
external_api
EcoLogitsGenAIExternalAPI the call is routed to.
An instance of EcoLogitsGenAIExternalAPI.
output_token_count
Average number of tokens generated per call. Drives generation latency, energy use, and embodied GWP through the EcoLogits impact model.
Unit: dimensionless.
data_stored
Unit: megabyte_stored.
Fixed by EcoLogitsGenAIExternalAPIJob to 0.0 MB_stored — not configurable.
compute_needed
Unit: cpu_core.
Fixed by EcoLogitsGenAIExternalAPIJob to 0.0 cpu_core — not configurable.
ram_needed
Unit: megabyte_ram.
Fixed by EcoLogitsGenAIExternalAPIJob to 0.0 MB_ram — not configurable.
Backwards links
Calculated attributes
data_transferred
Data transferred per call, estimated as 5 bytes per token (4 bytes UTF-8 plus 1 byte JSON overhead) times the output token count.
Example value: 5 kB
Depends directly on:
through the following calculations:
You can also visit the link to Data transferred for claude-opus-4-5’s full calculation graph.
impacts
Cached EcoLogits impact dictionary for one call, computed from the model parameters, output token count, and grid carbon intensity. Subsequent updates extract individual fields from this dictionary.
Example value: {'model_active_parameter_count': 133.5, 'model_total_paramet...
Depends directly on:
- claude-opus-4-5 active parameter count (in billions)
- claude-opus-4-5 total parameter count (in billions)
- Output token count for claude-opus-4-5
- Average carbon intensity of electricity mix for anthropic
- claude-opus-4-5 token per second
- claude-opus-4-5 time to first token
through the following calculations:
You can also visit the link to Ecologits impacts’s full calculation graph.
gpu_energy
Extracts the gpu_energy field from the cached EcoLogits impact dictionary on this job, converted into a typed e-footprint quantity.
Example value: 117 mWh
Depends directly on:
through the following calculations:
You can also visit the link to Ecologits gpu_energy for claude-opus-4-5’s full calculation graph.
generation_latency
Extracts the generation_latency field from the cached EcoLogits impact dictionary on this job, converted into a typed e-footprint quantity.
Example value: 30.9 s
Depends directly on:
through the following calculations:
You can also visit the link to Ecologits generation_latency for claude-opus-4-5’s full calculation graph.
model_required_memory
Extracts the model_required_memory field from the cached EcoLogits impact dictionary on this job, converted into a typed e-footprint quantity.
Example value: 1.61 TB ram
Depends directly on:
through the following calculations:
You can also visit the link to Ecologits model_required_memory for claude-opus-4-5’s full calculation graph.
gpu_required_count
Extracts the gpu_required_count field from the cached EcoLogits impact dictionary on this job, converted into a typed e-footprint quantity.
Example value: 32
Depends directly on:
through the following calculations:
You can also visit the link to Ecologits gpu_required_count for claude-opus-4-5’s full calculation graph.
server_energy
Extracts the server_energy field from the cached EcoLogits impact dictionary on this job, converted into a typed e-footprint quantity.
Example value: 644 mWh
Depends directly on:
through the following calculations:
You can also visit the link to Ecologits server_energy for claude-opus-4-5’s full calculation graph.
request_energy
Extracts the request_energy field from the cached EcoLogits impact dictionary on this job, converted into a typed e-footprint quantity.
Example value: 4.88 Wh
Depends directly on:
through the following calculations:
You can also visit the link to Ecologits request_energy for claude-opus-4-5’s full calculation graph.
request_usage_gwp
Extracts the request_usage_gwp field from the cached EcoLogits impact dictionary on this job, converted into a typed e-footprint quantity.
Example value: 1.87 g
Depends directly on:
through the following calculations:
You can also visit the link to Ecologits request_usage_gwp for claude-opus-4-5’s full calculation graph.
server_gpu_embodied_gwp
Extracts the server_gpu_embodied_gwp field from the cached EcoLogits impact dictionary on this job, converted into a typed e-footprint quantity.
Example value: 31.5 t
Depends directly on:
through the following calculations:
You can also visit the link to Ecologits server_gpu_embodied_gwp for claude-opus-4-5’s full calculation graph.
request_embodied_gwp
Extracts the request_embodied_gwp field from the cached EcoLogits impact dictionary on this job, converted into a typed e-footprint quantity.
Example value: 161 mg
Depends directly on:
through the following calculations:
You can also visit the link to Ecologits request_embodied_gwp for claude-opus-4-5’s full calculation graph.
request_duration
Request duration of one call, equal to the generation latency derived from EcoLogits.
Example value: 30.9 s
Depends directly on:
through the following calculations:
You can also visit the link to Ecologits generation_latency for claude-opus-4-5’s full calculation graph.
hourly_occurrences_per_usage_pattern
Hourly count of job invocations broken down by usage pattern, derived from when each usage pattern's journeys start and at what point in the journey this job is triggered.
Example value: {
UsagePattern usage pattern (60c504): 26298 values from 2025-01-01 00:00:00+00:00 to 2028-01-01 18:00:00+00:00 in :
first 10 vals [7, 2, 3, 1, 5, 7, 6, 4, 8, 5],
last 10 vals [6, 3, 3, 4, 1, 5, 2, 6, 3, 5],
}
Depends directly on:
through the following calculations:
You can also visit the link to Hourly occurrences in usage pattern’s full calculation graph.
hourly_avg_occurrences_per_usage_pattern
Hourly count of job invocations averaged with respect to job duration, so a job that runs longer than an hour contributes a fractional occurrence to several modeling buckets.
Example value: {
UsagePattern usage pattern (60c504): 26298 values from 2025-01-01 00:00:00+00:00 to 2028-01-01 18:00:00+00:00 in :
first 10 vals [0.0601, 0.0172, 0.0258, 0.00859, 0.0429, 0.0601, 0.0515, 0.0344, 0.0687, 0.0429],
last 10 vals [0.0515, 0.0258, 0.0258, 0.0344, 0.00859, 0.0429, 0.0172, 0.0515, 0.0258, 0.0429],
}
Depends directly on:
through the following calculations:
You can also visit the link to Average hourly occurrences in usage pattern’s full calculation graph.
hourly_data_transferred_per_usage_pattern
Hourly volume of data transferred over the network by this job, broken down by usage pattern.
Example value: {
UsagePattern usage pattern (60c504): 26298 values from 2025-01-01 00:00:00+00:00 to 2028-01-01 18:00:00+00:00 in kB:
first 10 vals [35, 10, 15, 5, 25, 35, 30, 20, 40, 25],
last 10 vals [30, 15, 15, 20, 5, 25, 10, 30, 15, 25],
}
Depends directly on:
- Average hourly occurrences in usage pattern
- Data transferred for claude-opus-4-5
- Ecologits generation_latency for claude-opus-4-5
through the following calculations:
You can also visit the link to Hourly data transferred in usage pattern’s full calculation graph.
hourly_data_stored_per_usage_pattern
Hourly net change in storage volume caused by this job, broken down by usage pattern.
Example value: {
UsagePattern usage pattern (60c504): 26298 values from 2025-01-01 00:00:00+00:00 to 2028-01-01 18:00:00+00:00 in B stored:
first 10 vals [0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
last 10 vals [0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
}
Depends directly on:
- Average hourly occurrences in usage pattern
- Data stored by request
- Ecologits generation_latency for claude-opus-4-5
through the following calculations:
You can also visit the link to Hourly data stored in usage pattern’s full calculation graph.
hourly_avg_occurrences_across_usage_patterns
Total hourly count of duration-averaged job invocations summed over every usage pattern.
Example value: 26298 values from 2025-01-01 00:00:00+00:00 to 2028-01-01 18:00:00+00:00 in :
first 10 vals [0.0601, 0.0172, 0.0258, 0.00859, 0.0429, 0.0601, 0.0515, 0.0344, 0.0687, 0.0429],
last 10 vals [0.0515, 0.0258, 0.0258, 0.0344, 0.00859, 0.0429, 0.0172, 0.0515, 0.0258, 0.0429]
Depends directly on:
through the following calculations:
You can also visit the link to Hourly average occurrences across usage patterns’s full calculation graph.
hourly_data_transferred_across_usage_patterns
Total hourly volume of data transferred over the network by this job, summed over every usage pattern.
Example value: 26298 values from 2025-01-01 00:00:00+00:00 to 2028-01-01 18:00:00+00:00 in kB:
first 10 vals [35, 10, 15, 5, 25, 35, 30, 20, 40, 25],
last 10 vals [30, 15, 15, 20, 5, 25, 10, 30, 15, 25]
Depends directly on:
through the following calculations:
You can also visit the link to Hourly data transferred across usage patterns’s full calculation graph.
hourly_data_stored_across_usage_patterns
Total hourly net change in storage volume caused by this job, summed over every usage pattern.
Example value: 26298 values from 2025-01-01 00:00:00+00:00 to 2028-01-01 18:00:00+00:00 in B stored:
first 10 vals [0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
last 10 vals [0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
Depends directly on:
through the following calculations:
You can also visit the link to Hourly data stored across usage patterns’s full calculation graph.
fabrication_impact_repartition_weights
Weights used to attribute fabrication-phase emissions of upstream impact sources to each container of this object.
Example value: {
UsageJourneyStep 20 min streaming (264451): 26298 values from 2025-01-01 00:00:00+00:00 to 2028-01-01 18:00:00+00:00 in :
first 10 vals [2.33, 0.667, 1, 0.333, 1.67, 2.33, 2, 1.33, 2.67, 1.67],
last 10 vals [2, 1, 1, 1.33, 0.333, 1.67, 0.667, 2, 1, 1.67],
}
Depends directly on:
through the following calculations:
You can also visit the link to 20 min streaming weight in impact repartition’s full calculation graph.
fabrication_impact_repartition_weight_sum
Sum of fabrication impact repartition weights, used as the denominator when normalising into per-container shares.
Example value: 26298 values from 2025-01-01 00:00:00+00:00 to 2028-01-01 18:00:00+00:00 in :
first 10 vals [2.33, 0.667, 1, 0.333, 1.67, 2.33, 2, 1.33, 2.67, 1.67],
last 10 vals [2, 1, 1, 1.33, 0.333, 1.67, 0.667, 2, 1, 1.67]
Depends directly on:
through the following calculations:
You can also visit the link to Fabrication impact repartition weights sum’s full calculation graph.
fabrication_impact_repartition
Normalised share of fabrication-phase emissions that this object attributes to each container.
Example value: {
UsageJourneyStep 20 min streaming (264451): 26298 values from 2025-01-01 00:00:00+00:00 to 2028-01-01 18:00:00+00:00 in :
first 10 vals [1, 1, 1, 1, 1, 1, 1, 1, 1, 1],
last 10 vals [1, 1, 1, 1, 1, 1, 1, 1, 1, 1],
}
Depends directly on:
through the following calculations:
You can also visit the link to fabrication impact attribution to 20 min streaming’s full calculation graph.
usage_impact_repartition_weights
Weights used to attribute usage-phase emissions of upstream impact sources to each container of this object.
Example value: {
UsageJourneyStep 20 min streaming (264451): 26298 values from 2025-01-01 00:00:00+00:00 to 2028-01-01 18:00:00+00:00 in ·g/kWh:
first 10 vals [198, 56.7, 85, 28.3, 142, 198, 170, 113, 227, 142],
last 10 vals [170, 85, 85, 113, 28.3, 142, 56.7, 170, 85, 142],
}
Depends directly on:
through the following calculations:
You can also visit the link to 20 min streaming weight in impact repartition’s full calculation graph.
usage_impact_repartition_weight_sum
Sum of usage impact repartition weights, used as the denominator when normalising into per-container shares.
Example value: 26298 values from 2025-01-01 00:00:00+00:00 to 2028-01-01 18:00:00+00:00 in ·g/kWh:
first 10 vals [198, 56.7, 85, 28.3, 142, 198, 170, 113, 227, 142],
last 10 vals [170, 85, 85, 113, 28.3, 142, 56.7, 170, 85, 142]
Depends directly on:
through the following calculations:
You can also visit the link to Usage impact repartition weights sum’s full calculation graph.
usage_impact_repartition
Normalised share of usage-phase emissions that this object attributes to each container.
Example value: {
UsageJourneyStep 20 min streaming (264451): 26298 values from 2025-01-01 00:00:00+00:00 to 2028-01-01 18:00:00+00:00 in :
first 10 vals [1, 1, 1, 1, 1, 1, 1, 1, 1, 1],
last 10 vals [1, 1, 1, 1, 1, 1, 1, 1, 1, 1],
}
Depends directly on:
through the following calculations:
You can also visit the link to usage impact attribution to 20 min streaming’s full calculation graph.
hourly_occurrences_across_usage_patterns
Hourly count of occurrences of this job summed across all usage patterns that trigger it.
Example value: 26298 values from 2025-01-01 00:00:00+00:00 to 2028-01-01 18:00:00+00:00 in :
first 10 vals [7, 2, 3, 1, 5, 7, 6, 4, 8, 5],
last 10 vals [6, 3, 3, 4, 1, 5, 2, 6, 3, 5]
Depends directly on:
through the following calculations:
You can also visit the link to Hourly occurrences across usage patterns’s full calculation graph.