nvidia/NVIDIA-Nemotron-Labs-3-Puzzle-75B-A9B-BF16
| benchmark | category | score | context | conf | evidence |
|---|---|---|---|---|---|
| RULER | long-context | 95.1 | RULER @ 256k | medium | | RULER @ 256k | 95.1 | 96.7 | |
| RULER | long-context | 94.2 | RULER @ 512k | medium | | RULER @ 512k | 94.2 | 95.7 | |
| RULER | long-context | 92.2 | RULER @ 1M | medium | | RULER @ 1M | 92.2 | 93.9 | |
| AIME 2025 | math | 89.7 | AIME25 (no tools) | medium | | AIME25 (no tools) | 89.7 | 92.2 | |
| MMLU-Pro | knowledge | 82.4 | MMLU-Pro | medium | | MMLU-Pro | 82.4 | 83.8 | |
| LiveCodeBench | coding | 81.1 | LiveCodeBench (v5 2024-07↔2024-12) | medium | | LiveCodeBench (v5 2024-07↔2024-12) | 81.1 | 82.1 | |
| GPQA | reasoning | 79.5 | GPQA (with tools) | medium | | GPQA (with tools) | 79.5 | 81.3 | |
| GPQA | reasoning | 78.6 | GPQA (no tools) | medium | | GPQA (no tools) | 78.6 | 80.5 | |
| MMLU-Pro | knowledge | 77.5 | MMLU-ProX (avg over langs) | medium | | MMLU-ProX (avg over langs) | 77.5 | 79.5 | |
| IFBench | instruction-following | 71.9 | IFBench (prompt) | medium | | IFBench (prompt) | 71.9 | 73.4 | |
| Chatbot Arena | chat | 68.6 | Arena-Hard-V2 | medium | | Arena-Hard-V2 | 68.6 | 72.8 | |
| SciCode | coding | 40.6 | SciCode (subtask) | medium | | SciCode (subtask) | 40.6 | 42.3 | |
| Terminal-Bench | agentic-coding | 24.0 | Terminal Bench (hard subset) | medium | | Terminal Bench (hard subset) | 24.0 | 25.5 | |
| Humanity's Last Exam | knowledge | 16.5 | HLE (no tools) | medium | | HLE (no tools) | 16.5 | 18.5 | |
none scraped.