Humanity's Last Exam with tools (HLE w/ tools)
Tool-augmented Humanity's Last Exam scores reported in DeepSeek-V4 thinking-mode evaluations.
Top models on HLE w/ tools — October 29, 2026
As of October 29, 2026, Kimi K3.1 Pro leads the HLE w/ tools leaderboard with 70.0% , followed by Claude Opus 5.5 (67.7%) and Claude Opus 5 (64.7%).
Kimi K3.1 Pro
Moonshot AI
Claude Opus 5.5
Anthropic
Claude Opus 5
Anthropic
36 modelsAgentic3% of Agentic reference weightCurrentUpdated October 29, 2026
Leaderboard (36 models)
| Rank | Model / configuration | Score | Parameters (B) | Open / closed |
|---|---|---|---|---|
| 1 | Kimi K3.1 ProMoonshot AI | 70.0% | Not reported | Pending |
| 2 | Claude Opus 5.5Anthropic | 67.7% | Not reported | Closed |
| 3 | Claude Opus 5Anthropic | 64.7% | Not reported | Closed |
| 4 | Claude Sonnet 5.5Anthropic | 64.5% | Not reported | Closed |
| 5 | DeepSeek V4.1 Flash 1017DeepSeek | 63.9% | Not reported | Open |
| 6 | GLM-5.3Z.AI | 62.5% | Not reported | Open |
| 7 | Muse Spark OpenMeta | 61.0% | Not reported | Open |
| 8 | DeepSeek V4 Pro 0813DeepSeek | 60.0% | Not reported | Open |
| 9 | Claude Sonnet 5Anthropic | 57.4% | Not reported | Closed |
| 10 | Claude Haiku 5.5Anthropic | 57.4% | Not reported | Closed |
| 11 | GPT-6 AstraOpenAI | 57.2% | Not reported | Closed |
| 12 | Ling 3.0 Flash SanteInclusionAI | 57.1% | Not reported | Open |
| 13 | Qwen3.8 MaxAlibaba | 56.2% | Not reported | Open |
| 14 | Apodex 1.1Apodex | 56.1% | Not reported | Closed |
| 15 | Ornith-1.5-397BOrnith AI | 56.1% | Not reported | Open |
| 16 | Hy4 previewTencent | 55.4% | Not reported | Open |
| 17 | GLM-5.3-FlashZ.AI | 55.3% | Not reported | Open |
| 18 | Sage-A1.5InternScience | 55.1% | Not reported | Open |
| 19 | GLM-5.2Z.AI | 54.7% | Not reported | Open |
| 20 | Qwen3.7 MaxAlibaba | 53.5% | Not reported | Closed |
| 21 | dots3-note PreviewDots Studio | 52.6% | Not reported | Open |
| 22 | GLM-5.1Z.AI | 52.3% | Not reported | Open |
| 23 | Beam 1.0 501B-A23B FP8Reflection AI | 51.1% | Not reported | Open |
| 24 | Beam 1.0 501B-A23B NVFP4Reflection AI | 51.1% | Not reported | Open |
| 25 | Beam 1.0 501B-A23BReflection AI | 51.1% | Not reported | Open |
| 26 | GLM-5Z.AI | 50.4% | Not reported | Open |
| 27 | Agents-A1InternScience | 47.6% | Not reported | Open |
| 28 | Step 3.7 FlashStepFun | 47.2% | Not reported | Open |
| 29 | DeepSeek V4 Flash 0731DeepSeek | 45.1% | Not reported | Open |
| 30 | DeepSeek V4 Pro (High)DeepSeek | 44.7% | Not reported | Open |
| 31 | DeepSeek V4 Flash (High)DeepSeek | 40.3% | Not reported | Open |
| 32 | Nemotron 3 UltraNVIDIA | 37.4% | Not reported | Open |
| 33 | Spark-X2.6-4BSparkLLM | 36.9% | Not reported | Open |
| 34 | Ornith-1.5-35B-A3BOrnith AI | 33.4% | Not reported | Open |
| 35 | Ornith-1.5-9BOrnith AI | 30.5% | Not reported | Open |
| 36 | LFM3-1.7BLiquidAI | 13.9% | Not reported | Open |
According to BenchLM.ai, Kimi K3.1 Pro leads the HLE w/ tools benchmark with a score of 70.0%, followed by Claude Opus 5.5 (67.7%) and Claude Opus 5 (64.7%). The scores show moderate spread, with meaningful differences between the top tier and mid-tier models.
36 models have been evaluated on HLE w/ tools. The benchmark falls in the Agentic category. BenchAlign v5.8 gives HLE w/ tools 3% of the Agentic reference weight, so it moves the Agentic leaderboard and the overall ranking. Reference weights are relative weights in the calibrated model, not fixed shares of a score.
About HLE w/ tools
Year
2026
Tasks
Expert questions with tool use
Format
Pass@1
Difficulty
Frontier tool-augmented reasoning
BenchLM stores HLE with tools as a provider-table row when a provider publishes exact values.
Freshness and provenance
Version
HLE w/ tools 2026
Refresh cadence
Quarterly
Staleness state
Current
Question availability
Public benchmark set
BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.
Questions
What does HLE w/ tools measure?
Tool-augmented Humanity's Last Exam scores reported in DeepSeek-V4 thinking-mode evaluations.
Which model scores highest on HLE w/ tools?
Kimi K3.1 Pro by Moonshot AI currently leads with a score of 70.0% on HLE w/ tools.
How many models are evaluated on HLE w/ tools?
36 AI models have published results on HLE w/ tools in the BenchLM catalog.
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 6,100+ subscribers.
One email each week. Unsubscribe anytime.