# LLM Chat Evaluation on ArenaHard Source: [LLM Chat Evaluation on ArenaHard]() Collection index: [llms.txt]() Data updated: 2026-06-02T10:46:50.441Z Evaluation of LLM chat performance using the ArenaHard dataset, 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 | LLM Chat Evaluation | | Dataset (including any stated split) | ArenaHard | | Metric | Accuracy | | Unit | PERCENT | | 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 - [Accuracy](): Higher is better; unit: PERCENT. ## Recorded results | Method | Accuracy | Result date | Report paper | | --- | --- | --- | --- | | Official Instruct Model | 49.2 | Not recorded | 6a1e640adbd7150506f8564a | | GLM-4.7 (Scaffold-only) | 17.32 | Not recorded | 6a1e640adbd7150506f8564a | | GPT-5.4 (High) | 14.34 | Not recorded | 6a1e640adbd7150506f8564a | | GLM-4.7 & ANDES (Ours) | 14.11 | Not recorded | 6a1e640adbd7150506f8564a | | Opus-4.7 (xHigh) | 10.21 | Not recorded | 6a1e640adbd7150506f8564a | | Opus-4.6 (1M) | 9 | Not recorded | 6a1e640adbd7150506f8564a | | Opus-4.6 | 7.78 | Not recorded | 6a1e640adbd7150506f8564a | | GPT-5.2 | 6.49 | Not recorded | 6a1e640adbd7150506f8564a | | Gemini-3.1-Pro | 6.42 | Not recorded | 6a1e640adbd7150506f8564a | | MiniMax-M2.5 | 2.18 | Not recorded | 6a1e640adbd7150506f8564a | | Base Model (SmolLM3-3B) | 0.42 | Not recorded | 6a1e640adbd7150506f8564a | | Kimi-K2-Thinking | 0.42 | Not recorded | 6a1e640adbd7150506f8564a | | MiniMax-M2.1 | 0.42 | Not recorded | 6a1e640adbd7150506f8564a | | GPT-5.1-Codex-Max | 0.42 | Not recorded | 6a1e640adbd7150506f8564a | | GLM-4.7 (OpenCode) | 0.42 | Not recorded | 6a1e640adbd7150506f8564a | | Sonnet-4.5 | 0.24 | Not recorded | 6a1e640adbd7150506f8564a | | Qwen3-Max | 0.21 | Not recorded | 6a1e640adbd7150506f8564a | ### Official Instruct Model Accuracy: 49.2. Result ID: 6a1eb47cfbc373592c19a23c. [Reporting paper]() Recorded evaluation configuration: {"Protocol":"Official Instruct","Backbone":"SmolLM3-3B"}. This record does not assert independent reproduction. ### GLM-4.7 (Scaffold-only) Accuracy: 17.32. Result ID: 6a1eb47cfbc373592c19a24b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Protocol":"Proposed Method","Backbone":"SmolLM3-3B"}. This record does not assert independent reproduction. ### GPT-5.4 (High) Accuracy: 14.34. Result ID: 6a1eb47cfbc373592c19a244. [Reporting paper]() Recorded evaluation configuration: {"Protocol":"Agent-Post-Trained","Backbone":"SmolLM3-3B"}. This record does not assert independent reproduction. ### GLM-4.7 & ANDES (Ours) Accuracy: 14.11. Result ID: 6a1eb47cfbc373592c19a24c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Protocol":"Proposed Method","Backbone":"SmolLM3-3B"}. This record does not assert independent reproduction. ### Opus-4.7 (xHigh) Accuracy: 10.21. Result ID: 6a1eb47cfbc373592c19a248. [Reporting paper]() Recorded evaluation configuration: {"Protocol":"Agent-Post-Trained","Backbone":"SmolLM3-3B"}. This record does not assert independent reproduction. ### Opus-4.6 (1M) Accuracy: 9. Result ID: 6a1eb47cfbc373592c19a249. [Reporting paper]() Recorded evaluation configuration: {"Protocol":"Agent-Post-Trained","Backbone":"SmolLM3-3B"}. This record does not assert independent reproduction. ### Opus-4.6 Accuracy: 7.78. Result ID: 6a1eb47cfbc373592c19a247. [Reporting paper]() Recorded evaluation configuration: {"Protocol":"Agent-Post-Trained","Backbone":"SmolLM3-3B"}. This record does not assert independent reproduction. ### GPT-5.2 Accuracy: 6.49. Result ID: 6a1eb47cfbc373592c19a245. [Reporting paper]() Recorded evaluation configuration: {"Protocol":"Agent-Post-Trained","Backbone":"SmolLM3-3B"}. This record does not assert independent reproduction. ### Gemini-3.1-Pro Accuracy: 6.42. Result ID: 6a1eb47cfbc373592c19a246. [Reporting paper]() Recorded evaluation configuration: {"Protocol":"Agent-Post-Trained","Backbone":"SmolLM3-3B"}. This record does not assert independent reproduction. ### MiniMax-M2.5 Accuracy: 2.18. Result ID: 6a1eb47cfbc373592c19a243. [Reporting paper]() Recorded evaluation configuration: {"Protocol":"Agent-Post-Trained","Backbone":"SmolLM3-3B"}. This record does not assert independent reproduction. ### Base Model (SmolLM3-3B) Accuracy: 0.42. Result ID: 6a1eb47cfbc373592c19a23d. [Reporting paper]() Recorded evaluation configuration: {"Protocol":"Zero-Shot","Backbone":"SmolLM3-3B"}. This record does not assert independent reproduction. ### Kimi-K2-Thinking Accuracy: 0.42. Result ID: 6a1eb47cfbc373592c19a240. [Reporting paper]() Recorded evaluation configuration: {"Protocol":"Agent-Post-Trained","Backbone":"SmolLM3-3B"}. This record does not assert independent reproduction. ### MiniMax-M2.1 Accuracy: 0.42. Result ID: 6a1eb47cfbc373592c19a241. [Reporting paper]() Recorded evaluation configuration: {"Protocol":"Agent-Post-Trained","Backbone":"SmolLM3-3B"}. This record does not assert independent reproduction. ### GPT-5.1-Codex-Max Accuracy: 0.42. Result ID: 6a1eb47cfbc373592c19a242. [Reporting paper]() Recorded evaluation configuration: {"Protocol":"Agent-Post-Trained","Backbone":"SmolLM3-3B"}. This record does not assert independent reproduction. ### GLM-4.7 (OpenCode) Accuracy: 0.42. Result ID: 6a1eb47cfbc373592c19a24a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Protocol":"Proposed Method","Backbone":"SmolLM3-3B"}. This record does not assert independent reproduction. ### Sonnet-4.5 Accuracy: 0.24. Result ID: 6a1eb47cfbc373592c19a23e. [Reporting paper]() Recorded evaluation configuration: {"Protocol":"Agent-Post-Trained","Backbone":"SmolLM3-3B"}. This record does not assert independent reproduction. ### Qwen3-Max Accuracy: 0.21. Result ID: 6a1eb47cfbc373592c19a23f. [Reporting paper]() Recorded evaluation configuration: {"Protocol":"Agent-Post-Trained","Backbone":"SmolLM3-3B"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.