Decoding Throughput on HumanEval
1,763.9Throughput (tok/s)FLARE-2B
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| FLARE-2BConcurrency level (C)=82026.06 | 1,763.9 | — | — | — | — | — | |
| LLaDA-2.1-miniConcurrency level (C)=82026.06 | 1,646.5 | — | — | — | — | — | |
| LLaDA-2.1-miniConcurrency level (C)=42026.06 | 1,471.4 | — | — | — | — | — | |
| FLARE-4BConcurrency level (C)=82026.06 | 1,178.9 | — | — | — | — | — | |
| FLARE-2BConcurrency level (C)=42026.06 | 1,113.3 | — | — | — | — | — | |
| FLARE-9BConcurrency level (C)=82026.06 | 1,007.1 | — | — | — | — | — | |
| FLARE-4BConcurrency level (C)=42026.06 | 812.2 | — | — | — | — | — | |
| FLARE-9BConcurrency level (C)=42026.06 | 642.8 | — | — | — | — | — | |
| LLaDA-2.0-miniConcurrency level (C)=82026.06 | 626.1 | — | — | — | — | — | |
| LLaDA-2.0-miniConcurrency level (C)=42026.06 | 496.7 | — | — | — | — | — | |
| LLaDA-2.1-miniConcurrency level (C)=12026.06 | 472.4 | — | — | — | — | — | |
| SDAR-1.7BConcurrency level (C)=82026.06 | 427 | — | — | — | — | — | |
| SDAR-4BConcurrency level (C)=82026.06 | 398.3 | — | — | — | — | — | |
| FLARE-2BConcurrency level (C)=12026.06 | 373.5 | — | — | — | — | — | |
| SDAR-8BConcurrency level (C)=82026.06 | 372.1 | — | — | — | — | — | |
| SDAR-1.7BConcurrency level (C)=42026.06 | 364.7 | — | — | — | — | — | |
| SDAR-4BConcurrency level (C)=42026.06 | 326.1 | — | — | — | — | — | |
| SDAR-8BConcurrency level (C)=42026.06 | 276.8 | — | — | — | — | — | |
| FLARE-4BConcurrency level (C)=12026.06 | 268.2 | — | — | — | — | — | |
| LLaDA-2.0-miniConcurrency level (C)=12026.06 | 234.2 | — | — | — | — | — | |
| SDAR-30B-A3BConcurrency level (C)=82026.06 | 214.1 | — | — | — | — | — | |
| FLARE-9BConcurrency level (C)=12026.06 | 203.2 | — | — | — | — | — | |
| SDAR-1.7BConcurrency level (C)=12026.06 | 135.6 | — | — | — | — | — | |
| SDAR-30B-A3BConcurrency level (C)=42026.06 | 122.3 | — | — | — | — | — | |
| SDAR-4BConcurrency level (C)=12026.06 | 105.1 | — | — | — | — | — | |
| SDAR-8BConcurrency level (C)=12026.06 | 83.8 | — | — | — | — | — | |
| SDAR-30B-A3BConcurrency level (C)=12026.06 | 39.8 | — | — | — | — | — | |
| AutoregressiveModel=Llama-3.1-Instruct 70B/1B2026.03 | — | 54.7 | — | — | — | — | |
| AutoregressiveModel=Qwen-3 32B/0.6B2026.03 | — | 88.8 | — | — | — | — | |
| Speculative DecodingModel=Llama-3.1-Instruct 70B/1B2026.03 | — | — | 176 | — | — | — | |
| Speculative DecodingModel=Qwen-3 32B/0.6B2026.03 | — | — | 146 | — | — | — | |
| SSDModel=Llama-3.1-Instruct 70B/1B2026.03 | — | — | — | 283 | 1.6 | 5.17 | |
| SSDModel=Qwen-3 32B/0.6B2026.03 | — | — | — | 222 | 1.52 | 2.5 |