Long-form Generation on WritingBench length-constrained
8.6L_R ScoreDeepSeek-V3.2-671B
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| DeepSeek-V3.2-671BType=Open-source LLM, Parameters=671B2026.06 | 8.6 | 8.59 | 8.53 | 8.55 | 8.59 | 8.73 | 8.63 | 8.5 | |
| IS-Writer-8BType=Capability-enhanced LLM, Parameters=8B2026.06 | 8.6 | 8.44 | 8.56 | 8.61 | 8.6 | 8.68 | 8.6 | 8.47 | |
| DeepSeek-R1-671BType=Open-source LLM, Parameters=671B2026.06 | 8.48 | 8.23 | 8.58 | 8.52 | 8.57 | 8.5 | 8.46 | 8.27 | |
| Gemini-2.5-flashType=Proprietary LLM2026.06 | 8.47 | 8.49 | 8.44 | 8.41 | 8.48 | 8.55 | 8.49 | 8.43 | |
| Qwen3-235B-A22B-InstructType=Open-source LLM, Parameters=235B2026.06 | 8.45 | 8.45 | 8.58 | 8.43 | 8.4 | 8.6 | 8.44 | 8.26 | |
| Writing-Model-Qwen-7BType=Capability-enhanced LLM, Parameters=7B2026.06 | 8.35 | 8.39 | 8.57 | 8.26 | 8.38 | 8.21 | 8.49 | 8.3 | |
| DeepSeek-V3-671BType=Open-source LLM, Parameters=671B2026.06 | 8.05 | 7.57 | 7.93 | 7.77 | 8.11 | 8.1 | 8.11 | 8.1 | |
| GPT-4o-miniType=Proprietary LLM2026.06 | 7.47 | 7.41 | 7.41 | 7.01 | 7.18 | 7.56 | 7.73 | 7.78 | |
| GPT-4oType=Proprietary LLM2026.06 | 7.35 | 7.16 | 7.39 | 6.86 | 7.03 | 7.45 | 7.48 | 7.7 | |
| Claude-3.5-sonnetType=Proprietary LLM2026.06 | 7.34 | 7.08 | 7.23 | 7.19 | 6.64 | 7.41 | 7.71 | 7.8 | |
| LLaMA3.3-70B-InstructType=Open-source LLM, Parameters=70B2026.06 | 6.58 | 6.4 | 6.52 | 6.1 | 6.26 | 6.43 | 7.13 | 7.01 | |
| Deepwriter-8BType=Capability-enhanced LLM, Parameters=8B2026.06 | 5.81 | 5.44 | 6.29 | 4.69 | 5.23 | 5.79 | 6.23 | 6.46 | |
| LongWriter-8BType=Capability-enhanced LLM, Parameters=8B2026.06 | 4.44 | 4.14 | 4.5 | 3.66 | 4.12 | 4.13 | 5.24 | 5 | |
| Suri-I-ORPO-7BType=Capability-enhanced LLM, Parameters=7B2026.06 | 2.24 | 2.05 | 2.45 | 2.7 | 1.83 | 2 | 2.41 | 2.43 |