Skip to main content
BenchLM
Data

Humanity's Last Exam with tools (HLE w/ tools)

Data verified

Tool-augmented Humanity's Last Exam scores reported in DeepSeek-V4 thinking-mode evaluations.

Top models on HLE w/ tools — November 18, 2026

As of November 18, 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%).

39 modelsAgentic3% of Agentic reference weightCurrentUpdated November 18, 2026

Leaderboard (39 models)

All 39 results on this page were reported by the labs that built the models.

HLE w/ tools leaderboard
RankModel / configurationScoreParameters (B)Open / closed
1Kimi K3.1 ProMoonshot AIReported by Moonshot AI
70.0%
Not reportedOpen
2Claude Opus 5.5AnthropicReported by Anthropic
67.7%
Not reportedClosed
3Claude Opus 5AnthropicReported by Anthropic
64.7%
Not reportedClosed
4Claude Sonnet 5.5AnthropicReported by Anthropic
64.5%
Not reportedClosed
5DeepSeek V4.1 Flash 1017DeepSeekReported by DeepSeek
63.9%
Not reportedOpen
6LongCat-2.5MeituanReported by Meituan
63.1%
Not reportedOpen
7GLM-5.3Z.AIReported by Z.AI
62.5%
Not reportedOpen
8Muse Spark OpenMetaReported by Meta
61.0%
Not reportedOpen
9DeepSeek V4 ProDeepSeekReported by DeepSeek
60.0%
Not reportedOpen
10Hy4TencentReported by Tencent
59.1%
Not reportedPending
11Claude Sonnet 5AnthropicReported by Anthropic
57.4%
Not reportedClosed
12Claude Haiku 5.5AnthropicReported by Anthropic
57.4%
Not reportedClosed
13GPT-6 AstraOpenAIReported by OpenAI
57.2%
Not reportedClosed
14Ling 3.0 Flash SanteInclusionAIReported by InclusionAI
57.1%
Not reportedOpen
15Qwen3.8 MaxAlibabaReported by Alibaba
56.2%
Not reportedOpen
16Apodex 1.1ApodexReported by Apodex
56.1%
Not reportedClosed
17Ornith-1.5-397BOrnith AIReported by Ornith AI
56.1%
Not reportedOpen
18Hy4 previewTencentReported by Tencent
55.4%
Not reportedOpen
19GLM-5.3-FlashZ.AIReported by Z.AI
55.3%
Not reportedOpen
20Hy4 flashTencentReported by Tencent
55.2%
Not reportedPending
21Sage-A1.5InternScienceReported by InternScience
55.1%
Not reportedOpen
22GLM-5.2Z.AIReported by Z.AI
54.7%
Not reportedOpen
23Qwen3.7 MaxAlibabaReported by Alibaba
53.5%
Not reportedClosed
24dots3-note PreviewDots StudioReported by Dots Studio
52.6%
Not reportedOpen
25GLM-5.1Z.AIReported by Z.AI
52.3%
Not reportedOpen
26Beam 1.0 501B-A23B FP8Reflection AIReported by Reflection AI
51.1%
Not reportedOpen
27Beam 1.0 501B-A23B NVFP4Reflection AIReported by Reflection AI
51.1%
Not reportedOpen
28Beam 1.0 501B-A23BReflection AIReported by Reflection AI
51.1%
Not reportedOpen
29GLM-5Z.AIReported by Z.AI
50.4%
Not reportedOpen
30Agents-A1InternScienceReported by InternScience
47.6%
Not reportedOpen
31Step 3.7 FlashStepFunReported by StepFun
47.2%
Not reportedOpen
32DeepSeek V4 Flash 0731DeepSeekReported by DeepSeek
45.1%
Not reportedOpen
33DeepSeek V4 Pro (High)DeepSeekReported by DeepSeek
44.7%
Not reportedOpen
34DeepSeek V4 Flash (High)DeepSeekReported by DeepSeek
40.3%
Not reportedOpen
35Nemotron 3 UltraNVIDIAReported by NVIDIA
37.4%
Not reportedOpen
36Spark-X2.6-4BSparkLLMReported by SparkLLM
36.9%
Not reportedOpen
37Ornith-1.5-35B-A3BOrnith AIReported by Ornith AI
33.4%
Not reportedOpen
38Ornith-1.5-9BOrnith AIReported by Ornith AI
30.5%
Not reportedOpen
39LFM3-1.7BLiquidAIReported by LiquidAI
13.9%
Not reportedOpen
  • Lab-reported
  • Benchmark-run
  • Independent
What the labels mean

Results reported by model developersUnranked

Labs ran these themselves, with a metric, split or setting the table above does not use. They take no rank here.

HLE w/ tools results reported by model developers, unranked
Apodex 1.1Apodex53.2ScoreReAct single agentThe stored row uses Agent Team; with toolsNot comparable to the scores aboveReported by Apodex
Step 3.7 FlashStepFun48.1ScoreModel card prose figureThe tool setup is not stated, and it differs from the stored launch-page valueNot comparable to the scores aboveReported by StepFun

A lab’s own result is shown for reference. It never takes a rank on this board.

How we label sources

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.

39 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

Current

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?

39 AI models have published results on HLE w/ tools in the BenchLM catalog.

Last updated: November 18, 2026 · BenchLM version HLE w/ tools 2026

Know when it’s worth switching models

The model to choose, the cheaper alternative, and the release we would wait on.

Read a sample issue

Join 7,000+ subscribers.

One email each week. Unsubscribe anytime.