Forge: leaderboard
Large Language Models (LLMs) have achieved remarkable success on reasoning benchmarks through Reinforcement Learning with Verifiable Rewards (RLVR), excelling at tasks such as math, coding, logic, and puzzles. However, existing benchmarks evaluate only correctness, while overlooking optimality, namely the ability to find the best solutions under constraints. We propose OPT-BENCH, the first compreh
Metric: Overall SR (self-reported). Source: benchmarklist.com. Saturation forecast: Estimated already saturated. 21 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | O3 | 91.5 | #124 |
| 2 | Qwen 3 235B A22B (Thinking) | 82.3 | #313 (Qwen 3 235B A22B) |
| 3 | Qwen 3 30B A3B | 76.8 | #493 |
| 4 | Qwen 3 32B | 70.7 | #443 |
| 5 | GPT-4o (2024-08-06) | 62.1 | #336 |
| 6 | Qwen 3 8B | 57.5 | #674 |
| 7 | Qwen 2.5 72B | 43.4 | #499 |
| 8 | Qwen 2.5 32B | 41.2 | #546 |
| 9 | Qwen 2.5 14B | 36.2 | #725 |
| 10 | Qwen 2.5 7B | 29.6 | #922 |
| 11 | Llama 3.1 8B | 15.1 | #1151 |
| 12 | Qwen 2.5 3B | 13.8 | #1211 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=forge · How It Works · Data refreshed daily, snapshot 2026-10-20.