LLM Stats leaderboard
37 scores · 37 models · first recorded Oct 2, 2026
higher is better · scale 0–1 · record Toolathlon
Best reported score
75.6%Muse Spark 1.1 · LLM Stats leaderboard
42 models tested · 2 sources · 11 source documents · latest report Aug 28, 2026
[ICLR 2026] The Tool Decathlon: Benchmarking Language Agents for Diverse, Realistic, and Long-Horizon Task Execution
Scores stay separate per source — never merged into one cross-source ranking.
37 scores · 37 models · first recorded Oct 2, 2026
higher is better · scale 0–1 · record Toolathlon
| Model | Value (%) | Reported | Score type | Recorded |
|---|---|---|---|---|
| GLM-5.3 | 73% | Aug 14, 2026 | unspecified | Oct 2, 2026 |
| DeepSeek-V4-Pro-0813 | 74.1% | Aug 13, 2026 | unspecified | Oct 2, 2026 |
| Qwen3.8 Max | 72.5% | Aug 2, 2026 | unspecified | Oct 2, 2026 |
| DeepSeek-V4-Flash-0731 | 70.3% | Jul 31, 2026 | unspecified | Oct 2, 2026 |
| Inkling-Small | 54.4% | Jul 30, 2026 | unspecified | Oct 2, 2026 |
| Laguna S 2.1 | 49.7% | Jul 21, 2026 | unspecified | Oct 2, 2026 |
| Kimi K3 | 73.2% | Jul 16, 2026 | unspecified | Oct 2, 2026 |
| Muse Spark 1.1 | 75.6% | Jul 9, 2026 | unspecified | Oct 2, 2026 |
| GPT-5.6 Sol | 58.0% | Jul 9, 2026 | unspecified | Oct 2, 2026 |
| GPT-5.6 Luna | 53.4% | Jul 9, 2026 | unspecified | Oct 2, 2026 |
| GPT-5.6 Terra | 53.1% | Jul 9, 2026 | unspecified | Oct 2, 2026 |
| Hy3 | 48.5% | Jul 6, 2026 | unspecified | Oct 2, 2026 |
| Claude Sonnet 5 | 54.3% | Jun 30, 2026 | unspecified | Oct 2, 2026 |
| Seed 2.1 Pro | 50.6% | Jun 24, 2026 | unspecified | Oct 2, 2026 |
| Seed 2.1 Turbo | 49.1% | Jun 24, 2026 | unspecified | Oct 2, 2026 |
| GLM-5.2 | 48.2% | Jun 16, 2026 | unspecified | Oct 2, 2026 |
| Claude Opus 4.8 | 59.9% | May 28, 2026 | unspecified | Oct 2, 2026 |
| Gemini 3.5 Flash | 56.5% | May 19, 2026 | unspecified | Oct 2, 2026 |
| GPT-5.5 | 55.6% | Apr 23, 2026 | unspecified | Oct 2, 2026 |
| DeepSeek-V4-Pro-Max | 51.8% | Apr 23, 2026 | unspecified | Oct 2, 2026 |
| DeepSeek-V4-Flash-Max | 47.8% | Apr 23, 2026 | unspecified | Oct 2, 2026 |
| DeepSeek-V4-Flash-0423 | 43.5% | Apr 23, 2026 | unspecified | Oct 2, 2026 |
| Kimi K2.6 | 50% | Apr 20, 2026 | unspecified | Oct 2, 2026 |
| Qwen3.6-35B-A3B | 26.9% | Apr 16, 2026 | unspecified | Oct 2, 2026 |
| GLM-5.1 | 40.7% | Apr 7, 2026 | unspecified | Oct 2, 2026 |
| Qwen3.6 Plus | 39.8% | Apr 2, 2026 | unspecified | Oct 2, 2026 |
| MiniMax M2.7 | 46.3% | Mar 18, 2026 | unspecified | Oct 2, 2026 |
| GPT-5.4 mini | 42.9% | Mar 17, 2026 | unspecified | Oct 2, 2026 |
| GPT-5.4 nano | 35.5% | Mar 17, 2026 | unspecified | Oct 2, 2026 |
| GPT-5.4 | 54.6% | Mar 5, 2026 | unspecified | Oct 2, 2026 |
| Qwen3.5-397B-A17B | 38.3% | Feb 16, 2026 | unspecified | Oct 2, 2026 |
| MiniMax M2.1 | 43.5% | Dec 23, 2025 | unspecified | Oct 2, 2026 |
| Gemini 3 Flash | 49.4% | Dec 17, 2025 | unspecified | Oct 2, 2026 |
| GPT-5.2 | 46.3% | Dec 11, 2025 | unspecified | Oct 2, 2026 |
| DeepSeek-V3.2 | 35.2% | Dec 1, 2025 | unspecified | Oct 2, 2026 |
| DeepSeek-V3.2 (Thinking) | 35.2% | Dec 1, 2025 | unspecified | Oct 2, 2026 |
| DeepSeek-V3.2-Speciale | 35.2% | Dec 1, 2025 | unspecified | Oct 2, 2026 |
10 scores · 10 models · first recorded Oct 2, 2026
higher is better · pass@1 · scale 0–100 · record Toolathlon Verified
| Model | Value (%) | Subset / variant | Test conditions | Reported by | Reported | Document |
|---|---|---|---|---|---|---|
| GLM-5.3-Flash | 78.4% | toolathlon_verified | official evaluation service, pass@1 averaged over 3 independent runs | Z.ai | Aug 27, 2026 | zai_glm_5_3_flash_model_card |
| Kimi K3 | 76.5% | tool_decathlon | reasoning=max | Moonshot AI | Jun 13, 2026 | moonshot_kimi_k3_model_card |
| Hy4 preview | 74.1% | toolathlon_verified | as printed in Table 3 (Large Models) | Tencent | Aug 28, 2026 | tencent_hy4_preview |
| DeepSeek-V4-Pro-0813 | 74.1% | tool_decathlon | max reasoning effort, temp 1.0, top_p 0.95 | DeepSeek | Aug 13, 2026 | deepseek_v4_pro_0813_model_card |
| DeepSeek-V4-Pro | 51.8% | tool_decathlon | think max, pass@1 | DeepSeek | Apr 22, 2026 | deepseek_v4_model_card |
| GLM-5.2 | 48.2% | tool_decathlon | official evaluation service, max_token 128K | Z.ai | Jun 16, 2026 | zai_glm_5_2_model_card |
| DeepSeek-V4-Flash | 47.8% | tool_decathlon | think max, pass@1 | DeepSeek | Jun 25, 2026 | deepseek_v4_technical_report |
| GLM-5.1 | 40.7% | tool_decathlon | as reported | Z.ai | Apr 3, 2026 | zai_glm_5_1_model_card |
| Qwen3.5-397B-A17B | 38.3% | tool_decathlon | thinking (temp 0.6, top_p 0.95, top_k 20) | Qwen | Feb 16, 2026 | qwen3_5_model_card |
| GLM-5 | 38% | tool_decathlon | as reported | Z.ai | Feb 11, 2026 | zai_glm_5_model_card |
| Model | Value (%) | Reported | Score type | Recorded |
|---|---|---|---|---|
| Hy4 preview | 74.1% | Aug 28, 2026 | self-reported | Oct 2, 2026 |
| GLM-5.3-Flash | 78.4% | Aug 27, 2026 | self-reported | Oct 2, 2026 |
| DeepSeek-V4-Pro-0813 | 74.1% | Aug 13, 2026 | self-reported | Oct 2, 2026 |
| DeepSeek-V4-Flash | 47.8% | Jun 25, 2026 |
every document that defines or reports scores
| other | DeepSeek | DeepSeek-V4-Pro-0813Aug 13, 2026 |
| tech report | DeepSeek | DeepSeek-V4Jun 25, 2026 |
| other | Moonshot AI | Kimi K3Jun 13, 2026 |
| other | Qwen | Qwen3.5-397B-A17BFeb 16, 2026 |
| tech report | Tencent | Hy4 previewAug 28, 2026 |
| other | Z.ai | GLM-5.1Apr 3, 2026 |
| other | Z.ai | GLM-5.2Jun 16, 2026 |
| other | Z.ai | GLM-5.3-FlashAug 27, 2026 |
| other | Z.ai | GLM-5Feb 11, 2026 |
| other | – | Toolathlon– |
| other | DeepSeek | DeepSeek-V4-ProApr 22, 2026 |
Records are located via the tracking corpus (ktwu01/benchmark-radar live dataset). Every row cites its source document.
| self-reported |
| Oct 2, 2026 |
| GLM-5.2 | 48.2% | Jun 16, 2026 | self-reported | Oct 2, 2026 |
| Kimi K3 | 76.5% | Jun 13, 2026 | self-reported | Oct 2, 2026 |
| DeepSeek-V4-Pro | 51.8% | Apr 22, 2026 | self-reported | Oct 2, 2026 |
| GLM-5.1 | 40.7% | Apr 3, 2026 | self-reported | Oct 2, 2026 |
| Qwen3.5-397B-A17B | 38.3% | Feb 16, 2026 | self-reported | Oct 2, 2026 |
| GLM-5 | 38% | Feb 11, 2026 | self-reported | Oct 2, 2026 |