# Large Language Model Training Efficiency on Llama 1.7B 3.2

Source: [Large Language Model Training Efficiency on Llama 1.7B 3.2](<https://www.sota2.com/research/sota/large-language-model-training-efficiency-on-llama-1-7b-3-2>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-02-26T05:42:03.655Z

Evaluation of the training efficiency for the Llama 1.7B 3.2 model, specifically measuring reductions in energy consumption under iso-time constraints and time reduction under iso-energy constraints.

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

| Scope | Recorded value |
| --- | --- |
| Task | Large Language Model Training Efficiency |
| Dataset (including any stated split) | Llama 1.7B 3.2 |
| Metric | Energy Reduction (Iso-Time) |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 11 of 11 |
| Coverage | All results currently returned for this benchmark |

Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol.

## Metrics

- [Energy Reduction (Iso-Time)](<https://www.sota2.com/research/sota/large-language-model-training-efficiency-on-llama-1-7b-3-2.md?metric=Energy%20Reduction%20(Iso-Time)>): Higher is better; unit: PERCENT.

- [Time Reduction (Iso-Energy)](<https://www.sota2.com/research/sota/large-language-model-training-efficiency-on-llama-1-7b-3-2.md?metric=Time%20Reduction%20(Iso-Energy)>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Energy Reduction (Iso-Time) | Result date | Report paper |
| --- | --- | --- | --- |
| Kareus | 28.3 | Not recorded | 6978467ca2b28e761d87c477 |
| Kareus | 26.8 | Not recorded | 6978467ca2b28e761d87c477 |
| Kareus | 23.1 | Not recorded | 6978467ca2b28e761d87c477 |
| Nanobatching + Perseus | 20.4 | Not recorded | 6978467ca2b28e761d87c477 |
| Kareus | 20.1 | Not recorded | 6978467ca2b28e761d87c477 |
| Nanobatching + Perseus | 20 | Not recorded | 6978467ca2b28e761d87c477 |
| Nanobatching + Perseus | 16.8 | Not recorded | 6978467ca2b28e761d87c477 |
| Kareus | 15 | Not recorded | 6978467ca2b28e761d87c477 |
| Nanobatching + Perseus | 8.9 | Not recorded | 6978467ca2b28e761d87c477 |
| Kareus | 0 | Not recorded | 6978467ca2b28e761d87c477 |
| Nanobatching + Perseus | -0.7 | Not recorded | 6978467ca2b28e761d87c477 |

### Kareus

Energy Reduction (Iso-Time): 28.3. Result ID: 69991afb5698665ec4818d53.

[Reporting paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

Recorded evaluation configuration: {"Parallelism Strategy":"TP8","Number of Microbatches":16,"Sequence Length":"4K"}. This record does not assert independent reproduction.

### Kareus

Energy Reduction (Iso-Time): 26.8. Result ID: 69991afb5698665ec4818d4f.

[Reporting paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

Recorded evaluation configuration: {"Parallelism Strategy":"TP8","Number of Microbatches":8,"Sequence Length":"4K"}. This record does not assert independent reproduction.

### Kareus

Energy Reduction (Iso-Time): 23.1. Result ID: 69991afb5698665ec4818d51.

[Reporting paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

Recorded evaluation configuration: {"Parallelism Strategy":"TP8","Number of Microbatches":8,"Sequence Length":"8K"}. This record does not assert independent reproduction.

### Nanobatching + Perseus

Energy Reduction (Iso-Time): 20.4. Result ID: 69991afb5698665ec4818d52.

[Reporting paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

Recorded evaluation configuration: {"Parallelism Strategy":"TP8","Number of Microbatches":16,"Sequence Length":"4K"}. This record does not assert independent reproduction.

### Kareus

Energy Reduction (Iso-Time): 20.1. Result ID: 69991afb5698665ec4818d58.

[Reporting paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

Recorded evaluation configuration: {"Parallelism Strategy":"CP2TP4","Number of Microbatches":16,"Sequence Length":"4K"}. This record does not assert independent reproduction.

### Nanobatching + Perseus

Energy Reduction (Iso-Time): 20. Result ID: 69991afb5698665ec4818d50.

[Reporting paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

Recorded evaluation configuration: {"Parallelism Strategy":"TP8","Number of Microbatches":8,"Sequence Length":"8K"}. This record does not assert independent reproduction.

### Nanobatching + Perseus

Energy Reduction (Iso-Time): 16.8. Result ID: 69991afb5698665ec4818d4e.

[Reporting paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

Recorded evaluation configuration: {"Parallelism Strategy":"TP8","Number of Microbatches":8,"Sequence Length":"4K"}. This record does not assert independent reproduction.

### Kareus

Energy Reduction (Iso-Time): 15. Result ID: 69991afb5698665ec4818d56.

[Reporting paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

Recorded evaluation configuration: {"Parallelism Strategy":"CP2TP4","Number of Microbatches":8,"Sequence Length":"8K"}. This record does not assert independent reproduction.

### Nanobatching + Perseus

Energy Reduction (Iso-Time): 8.9. Result ID: 69991afb5698665ec4818d57.

[Reporting paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

Recorded evaluation configuration: {"Parallelism Strategy":"CP2TP4","Number of Microbatches":16,"Sequence Length":"4K"}. This record does not assert independent reproduction.

### Kareus

Energy Reduction (Iso-Time): 0. Result ID: 69991afb5698665ec4818d54.

[Reporting paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

Recorded evaluation configuration: {"Parallelism Strategy":"CP2TP4","Number of Microbatches":8,"Sequence Length":"4K"}. This record does not assert independent reproduction.

### Nanobatching + Perseus

Energy Reduction (Iso-Time): -0.7. Result ID: 69991afb5698665ec4818d55.

[Reporting paper](<https://www.sota2.com/research/paper/6978467ca2b28e761d87c477.md>)

Recorded evaluation configuration: {"Parallelism Strategy":"CP2TP4","Number of Microbatches":8,"Sequence Length":"8K"}. This record does not assert independent reproduction.

Page 1 of 1. This is an index page, not the full collection.
