Agents · Benchmark profile
Tool-augmented Humanity's Last Exam scores reported in DeepSeek-V4 thinking-mode evaluations.
Data verified 21 Jul 2026 · Methodology 1.6.0
Benchmark score on Humanity's Last Exam with tools
Claude Sonnet 5 leads at 57.4%, followed by Qwen3.7-Max (53.5%) and DeepSeek V4 Pro (Max) (48.2%).
Visual analysis
Switch between model placement, score distribution and descriptive provider averages. Every view uses the same sourced leaderboard.
9 ranked models · higher is better
| Model | Rank | Provider | Score |
|---|---|---|---|
| Claude Sonnet 5 | #1 | Anthropic | 57.4% |
| Qwen3.7-Max | #2 | Alibaba Cloud | 53.5% |
| DeepSeek V4 Pro (Max) | #3 | DeepSeek | 48.2% |
| Agents-A1 | #4 | InternScience | 47.6% |
| Step 3.7 Flash | #5 | StepFun | 47.2% |
| DeepSeek V4 Flash (Max) | #6 | DeepSeek | 45.1% |
| DeepSeek V4 Pro (High) | #7 | DeepSeek | 44.7% |
| DeepSeek V4 Flash (High) | #8 | DeepSeek | 40.3% |
| Nemotron 3 Ultra | #9 | NVIDIA | 37.4% |
One best score per model · higher is better
| Rank | Model | Provider | License | Evidence use | Score |
|---|---|---|---|---|---|
| #1 | Claude Sonnet 5 claude-sonnet-5 | Anthropic | closed | Estimated reference | 57.4% |
| #2 | Qwen3.7-Max qwen3.7-max | Alibaba Cloud | — | Reference only | 53.5% |
| #3 | DeepSeek V4 Pro (Max) deepseek-v4-pro-max | DeepSeek | open | Estimated reference | 48.2% |
| #4 | Agents-A1 agents-a1 | InternScience | open | Estimated reference | 47.6% |
| #5 | Step 3.7 Flash step-3-7-flash | StepFun | open | Estimated reference | 47.2% |
| #6 | DeepSeek V4 Flash (Max) deepseek-v4-flash-max | DeepSeek | open | Estimated reference | 45.1% |
| #7 | DeepSeek V4 Pro (High) deepseek-v4-pro-high | DeepSeek | open | Estimated reference | 44.7% |
| #8 | DeepSeek V4 Flash (High) deepseek-v4-flash-high | DeepSeek | open | Estimated reference | 40.3% |
| #9 | Nemotron 3 Ultra nemotron-3-ultra-550b | NVIDIA | open | Estimated reference | 37.4% |
The top of this snapshot is led by Claude Sonnet 5 at 57.4%; third place is 9.2 points behind. The top-9 spread is 20.0 points.
About Humanity's Last Exam with tools
Tool-augmented Humanity's Last Exam scores reported in DeepSeek-V4 thinking-mode evaluations. Results stay tied to the exact model variant and evaluation system. Multiple systems for the same model use the best published score on this page; overall Lumina scoring uses the median of ranking-eligible rows.
Open benchmark source ↗FAQ
Tool-augmented Humanity's Last Exam scores reported in DeepSeek-V4 thinking-mode evaluations.
Claude Sonnet 5 by Anthropic currently leads with 57.4%.
9 models in the LuminaBench cohort have a qualifying score on this benchmark.
No. This benchmark is display-only and does not enter the overall Lumina composite.
Related
