# Long-Context Inference on LongBench \>128K tokens v2 (long)

Source: [Long-Context Inference on LongBench \>128K tokens v2 (long)](<https://www.sota2.com/research/sota/long-context-inference-on-longbench-128k-tokens-v2-long>)

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

Data updated: 2026-02-26T05:33:25.118Z

Evaluation of language models on the LongBench v2 dataset, specifically focusing on tasks requiring context longer than 128K tokens, measured by Accuracy.

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 | Long-Context Inference |
| Dataset (including any stated split) | LongBench \>128K tokens v2 (long) |
| Metric | Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 8 of 8 |
| 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

- [Accuracy](<https://www.sota2.com/research/sota/long-context-inference-on-longbench-128k-tokens-v2-long.md?metric=Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| SnapKV (Best) | 35.2 | Not recorded | 699e81ed6017c4460ffacda4 |
| CHESS (Aggressive) | 33.3 | Not recorded | 699e81ed6017c4460ffacda4 |
| KeyDiff (Best) | 32.4 | Not recorded | 699e81ed6017c4460ffacda4 |
| H2O (Best) | 31.5 | Not recorded | 699e81ed6017c4460ffacda4 |
| Quest (Best) | 31.5 | Not recorded | 699e81ed6017c4460ffacda4 |
| CHESS (Moderate) | 31.5 | Not recorded | 699e81ed6017c4460ffacda4 |
| CHESS (Conservative) | 29.6 | Not recorded | 699e81ed6017c4460ffacda4 |
| FullKV Inference | 28.7 | Not recorded | 699e81ed6017c4460ffacda4 |

### SnapKV (Best)

Accuracy: 35.2. Result ID: 699eb6615698665ec484db71.

[Reporting paper](<https://www.sota2.com/research/paper/699e81ed6017c4460ffacda4.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73b21e209131cea5df89.md>)

[Implementation](<https://github.com/FasterDecoding/SnapKV>)

Recorded evaluation configuration: {"KV Cache Budget":"4096 tokens"}. This record does not assert independent reproduction.

### CHESS (Aggressive)

Accuracy: 33.3. Result ID: 699eb6615698665ec484db75.

[Reporting paper](<https://www.sota2.com/research/paper/699e81ed6017c4460ffacda4.md>)

[Method origin paper](<https://www.sota2.com/research/paper/699e81ed6017c4460ffacda4.md>)

Recorded evaluation configuration: {"KV Cache Budget":"1%"}. This record does not assert independent reproduction.

### KeyDiff (Best)

Accuracy: 32.4. Result ID: 699eb6615698665ec484db70.

[Reporting paper](<https://www.sota2.com/research/paper/699e81ed6017c4460ffacda4.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78a81e209131cea79ade.md>)

Recorded evaluation configuration: {"KV Cache Budget":"8192 tokens"}. This record does not assert independent reproduction.

### H2O (Best)

Accuracy: 31.5. Result ID: 699eb6615698665ec484db6f.

[Reporting paper](<https://www.sota2.com/research/paper/699e81ed6017c4460ffacda4.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa718c1e209131cea51ead.md>)

[Implementation](<https://github.com/FMInference/H2O>)

Recorded evaluation configuration: {"KV Cache Budget":"20%"}. This record does not assert independent reproduction.

### Quest (Best)

Accuracy: 31.5. Result ID: 699eb6615698665ec484db72.

[Reporting paper](<https://www.sota2.com/research/paper/699e81ed6017c4460ffacda4.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74a51e209131cea62e6e.md>)

Recorded evaluation configuration: {"KV Cache Budget":"2048 tokens"}. This record does not assert independent reproduction.

### CHESS (Moderate)

Accuracy: 31.5. Result ID: 699eb6615698665ec484db74.

[Reporting paper](<https://www.sota2.com/research/paper/699e81ed6017c4460ffacda4.md>)

[Method origin paper](<https://www.sota2.com/research/paper/699e81ed6017c4460ffacda4.md>)

Recorded evaluation configuration: {"KV Cache Budget":"40%"}. This record does not assert independent reproduction.

### CHESS (Conservative)

Accuracy: 29.6. Result ID: 699eb6615698665ec484db73.

[Reporting paper](<https://www.sota2.com/research/paper/699e81ed6017c4460ffacda4.md>)

[Method origin paper](<https://www.sota2.com/research/paper/699e81ed6017c4460ffacda4.md>)

Recorded evaluation configuration: {"KV Cache Budget":"73%"}. This record does not assert independent reproduction.

### FullKV Inference

Accuracy: 28.7. Result ID: 699eb6615698665ec484db6e.

[Reporting paper](<https://www.sota2.com/research/paper/699e81ed6017c4460ffacda4.md>)

Recorded evaluation configuration: {"KV Cache Budget":"100%"}. This record does not assert independent reproduction.

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