# LLM Training Performance on Qwen models on A6000 (48GB VRAM) 2.5 (train)

Source: [LLM Training Performance on Qwen models on A6000 (48GB VRAM) 2.5 (train)](<https://www.sota2.com/research/sota/llm-training-performance-on-qwen-models-on-a6000-48gb-vram-2-5-train>)

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

Data updated: 2026-04-08T07:50:18.283Z

Evaluation of LLM training performance metrics, including TFLOPS, GPU/CPU memory utilization, and token throughput, for various Qwen models running under the constraint of 48GB A6000 VRAM.

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 | LLM Training Performance |
| Dataset (including any stated split) | Qwen models on A6000 (48GB VRAM) 2.5 (train) |
| Metric | TFLOPS |
| Unit | SCALAR |
| Direction | Higher is better |
| Results loaded | 5 of 5 |
| 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

- [TFLOPS](<https://www.sota2.com/research/sota/llm-training-performance-on-qwen-models-on-a6000-48gb-vram-2-5-train.md?metric=TFLOPS>): Higher is better; unit: SCALAR.

- [GPU Memory Usage (GB)](<https://www.sota2.com/research/sota/llm-training-performance-on-qwen-models-on-a6000-48gb-vram-2-5-train.md?metric=GPU%20Memory%20Usage%20(GB)>): Lower is better; unit: SIZE.

- [CPU Memory Usage (GB)](<https://www.sota2.com/research/sota/llm-training-performance-on-qwen-models-on-a6000-48gb-vram-2-5-train.md?metric=CPU%20Memory%20Usage%20(GB)>): Lower is better; unit: SIZE.

- [Throughput (tokens/s)](<https://www.sota2.com/research/sota/llm-training-performance-on-qwen-models-on-a6000-48gb-vram-2-5-train.md?metric=Throughput%20(tokens%2Fs)>): Higher is better; unit: SCALAR.

## Recorded results

| Method | TFLOPS | Result date | Report paper |
| --- | --- | --- | --- |
| MegaTrain | 56.82 | Not recorded | 69d5e1247d0172140ec75b40 |
| MegaTrain | 55.73 | Not recorded | 69d5e1247d0172140ec75b40 |
| MegaTrain | 49.7 | Not recorded | 69d5e1247d0172140ec75b40 |
| ZeRO-3 Offload | 27.55 | Not recorded | 69d5e1247d0172140ec75b40 |
| ZeRO-3 Offload | 23.89 | Not recorded | 69d5e1247d0172140ec75b40 |

### MegaTrain

TFLOPS: 56.82. Result ID: 69d608b047f34ce924be578b.

[Reporting paper](<https://www.sota2.com/research/paper/69d5e1247d0172140ec75b40.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69d5e1247d0172140ec75b40.md>)

[Implementation](<https://github.com/DLYuanGod/MegaTrain>)

Recorded evaluation configuration: {"Model":"Qwen2.5-14B","Max BS":"9"}. This record does not assert independent reproduction.

### MegaTrain

TFLOPS: 55.73. Result ID: 69d608b047f34ce924be5789.

[Reporting paper](<https://www.sota2.com/research/paper/69d5e1247d0172140ec75b40.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69d5e1247d0172140ec75b40.md>)

[Implementation](<https://github.com/DLYuanGod/MegaTrain>)

Recorded evaluation configuration: {"Model":"Qwen2.5-7B","Max BS":"12"}. This record does not assert independent reproduction.

### MegaTrain

TFLOPS: 49.7. Result ID: 69d608b047f34ce924be5787.

[Reporting paper](<https://www.sota2.com/research/paper/69d5e1247d0172140ec75b40.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69d5e1247d0172140ec75b40.md>)

[Implementation](<https://github.com/DLYuanGod/MegaTrain>)

Recorded evaluation configuration: {"Model":"Qwen2.5-3B","Max BS":"15"}. This record does not assert independent reproduction.

### ZeRO-3 Offload

TFLOPS: 27.55. Result ID: 69d608b047f34ce924be578a.

[Reporting paper](<https://www.sota2.com/research/paper/69d5e1247d0172140ec75b40.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa678f1e209131cea17034.md>)

Recorded evaluation configuration: {"Model":"Qwen2.5-7B","Max BS":"1"}. This record does not assert independent reproduction.

### ZeRO-3 Offload

TFLOPS: 23.89. Result ID: 69d608b047f34ce924be5788.

[Reporting paper](<https://www.sota2.com/research/paper/69d5e1247d0172140ec75b40.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa678f1e209131cea17034.md>)

Recorded evaluation configuration: {"Model":"Qwen2.5-3B","Max BS":"1"}. This record does not assert independent reproduction.

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