# Triton kernel generation on KernelBench LEVEL3 1.0 Source: [Triton kernel generation on KernelBench LEVEL3 1.0]() Collection index: [llms.txt]() Data updated: 2026-02-26T05:45:12.340Z Evaluation of Triton kernel generation performance using the KernelBench LEVEL3 version 1.0 dataset, measured by Fast1, Fast1.2, Fast1.5, and Fast2 indicators. 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 | Triton kernel generation | | Dataset (including any stated split) | KernelBench LEVEL3 1.0 | | Metric | Fast1 Score | | Unit | SCALAR | | Direction | Higher is better | | Results loaded | 17 of 17 | | 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 - [Fast1 Score](): Higher is better; unit: SCALAR. - [Fast1.2 Score](): Higher is better; unit: SCALAR. - [Fast1.5 Score](): Higher is better; unit: SCALAR. - [Fast2 Score](): Higher is better; unit: SCALAR. ## Recorded results | Method | Fast1 Score | Result date | Report paper | | --- | --- | --- | --- | | DR. KERNEL-14B-STTS | 29.8 | Not recorded | 698575d8c97a0b3e3b5d5e92 | | GPT-5 | 21 | Not recorded | 698575d8c97a0b3e3b5d5e92 | | Claude-4.5-Sonnet | 21 | Not recorded | 698575d8c97a0b3e3b5d5e92 | | DR. KERNEL-14B-STTS | 17.1 | Not recorded | 698575d8c97a0b3e3b5d5e92 | | GPT-5 | 14 | Not recorded | 698575d8c97a0b3e3b5d5e92 | | Claude-4.5-Sonnet | 12 | Not recorded | 698575d8c97a0b3e3b5d5e92 | | DR. KERNEL-8B | 10.8 | Not recorded | 698575d8c97a0b3e3b5d5e92 | | DR. KERNEL-14B | 9.2 | Not recorded | 698575d8c97a0b3e3b5d5e92 | | DR. KERNEL-14B | 8.8 | Not recorded | 698575d8c97a0b3e3b5d5e92 | | AutoTriton | 7.5 | Not recorded | 698575d8c97a0b3e3b5d5e92 | | DR. KERNEL-8B | 7.2 | Not recorded | 698575d8c97a0b3e3b5d5e92 | | Qwen3-Coder-A30BA3 | 7 | Not recorded | 698575d8c97a0b3e3b5d5e92 | | Qwen3-8B | 5.7 | Not recorded | 698575d8c97a0b3e3b5d5e92 | | GLM-4.7 | 5 | Not recorded | 698575d8c97a0b3e3b5d5e92 | | Qwen3-32B | 3.5 | Not recorded | 698575d8c97a0b3e3b5d5e92 | | Deepseek-V3.2-Thinking | 2 | Not recorded | 698575d8c97a0b3e3b5d5e92 | | Cold-Start-8B | 0.5 | Not recorded | 698575d8c97a0b3e3b5d5e92 | ### DR. KERNEL-14B-STTS Fast1 Score: 29.8. Result ID: 6998ee085698665ec4529ca9. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"scaling\_method":"sequential test-time scaling (STTS)","selection\_strategy":"best turn selection"}. This record does not assert independent reproduction. ### GPT-5 Fast1 Score: 21. Result ID: 6998ee085698665ec4529c9d. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Claude-4.5-Sonnet Fast1 Score: 21. Result ID: 6998ee085698665ec4529c9e. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### DR. KERNEL-14B-STTS Fast1 Score: 17.1. Result ID: 6998ee085698665ec4529ca8. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"scaling\_method":"sequential test-time scaling (STTS)"}. This record does not assert independent reproduction. ### GPT-5 Fast1 Score: 14. Result ID: 6998ee085698665ec4529c7f. [Reporting paper]() Recorded evaluation configuration: {"Evaluation Engine":"torch.compile"}. This record does not assert independent reproduction. ### Claude-4.5-Sonnet Fast1 Score: 12. Result ID: 6998ee085698665ec4529c80. [Reporting paper]() Recorded evaluation configuration: {"Evaluation Engine":"torch.compile"}. This record does not assert independent reproduction. ### DR. KERNEL-8B Fast1 Score: 10.8. Result ID: 6998ee085698665ec4529ca6. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### DR. KERNEL-14B Fast1 Score: 9.2. Result ID: 6998ee085698665ec4529c82. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Evaluation Engine":"torch.compile"}. This record does not assert independent reproduction. ### DR. KERNEL-14B Fast1 Score: 8.8. Result ID: 6998ee085698665ec4529ca7. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### AutoTriton Fast1 Score: 7.5. Result ID: 6998ee085698665ec4529ca4. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### DR. KERNEL-8B Fast1 Score: 7.2. Result ID: 6998ee085698665ec4529c81. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Evaluation Engine":"torch.compile"}. This record does not assert independent reproduction. ### Qwen3-Coder-A30BA3 Fast1 Score: 7. Result ID: 6998ee085698665ec4529ca3. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Qwen3-8B Fast1 Score: 5.7. Result ID: 6998ee085698665ec4529ca1. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### GLM-4.7 Fast1 Score: 5. Result ID: 6998ee085698665ec4529ca0. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Qwen3-32B Fast1 Score: 3.5. Result ID: 6998ee085698665ec4529ca2. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Deepseek-V3.2-Thinking Fast1 Score: 2. Result ID: 6998ee085698665ec4529c9f. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Cold-Start-8B Fast1 Score: 0.5. Result ID: 6998ee085698665ec4529ca5. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"training":"cold-start data"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.