| Step 5 Preview | | 80.5 | | | Step 5 Preview (High) | 2026-09-20 |
| Kimi K3: Open Frontier Intelligence | ✓ Link | 77.8 | | | Kimi K3 (Max) (Kimi Code) | 2026-07-27 |
| ProgramBench Leaderboard | ✓ Link | 74.7 | 4.5 | 37.0 | Claude Opus 5 (xhigh) (mini-SWE-agent) | 2026-07-24 |
| ProgramBench Leaderboard | | 70.9 | 0.0 | 16.5 | Claude Opus 4.8 (xhigh) (mini-SWE-agent) | 2026-05-28 |
| Introducing GPT-5.5 | | 70.8 | | | GPT-5.5 (Max) (Claude Code) | 2026-04-23 |
| ProgramBench Leaderboard | | 70.8 | 2.5 | 25.0 | Muse Spark 1.3 (max) (mini-SWE-agent) | 2026-09-02 |
| ProgramBench Leaderboard | ✓ Link | 69.9 | 1.0 | 15.5 | GPT-5.6 Sol (xhigh) (mini-SWE-agent) | 2026-07-09 |
| ProgramBench Leaderboard | | 68.6 | 1.0 | 16.5 | Muse Spark 1.3 (xhigh) (mini-SWE-agent) | 2026-09-02 |
| GLM-5.2: Built for Long-Horizon Tasks | ✓ Link | 63.7 | | | GLM-5.2 (Claude Code) | 2026-06-16 |
| ProgramBench Leaderboard | ✓ Link | 61.2 | 0.0 | 5.5 | Gemini 3.7 Flash (mini-SWE-agent) | 2026-08-13 |
| RecreationWorld: Scalable and Verifiable Environments for Hybrid Computer-Use Agents | ✓ Link | 58.2 | | | Qwen-Flash-CPT (Claude Code) | 2026-09-18 |
| ProgramBench Leaderboard | ✓ Link | 57.8 | 0.5 | 2.5 | GPT-5.6 Sol (mini-SWE-agent) | 2026-07-09 |
| ProgramBench Leaderboard | | 57.2 | 0.0 | 4.5 | Muse Spark 1.2 (xhigh) (mini-SWE-agent) | 2026-08-05 |
| ProgramGym: Scaling Training Environments for Long-Horizon Program Reproduction Agents | | 56.25 | | | Qwen3.8-27B-SFT + Program-Agent | 2026-10-09 |
| ProgramBench Leaderboard | | 55.7 | 0.5 | 4.0 | Gemini 3.6 Flash (mini-SWE-agent) | 2026-07-21 |
| ProgramBench Leaderboard | | 55.1 | 0.0 | 4.5 | Claude Opus 4.7 (xhigh) (mini-SWE-agent) | 2026-04-16 |
| ProgramBench Leaderboard | | 53.6 | 0.0 | 3.0 | Gemini 3.5 Flash (mini-SWE-agent) | 2026-05-19 |
| Kimi K2.7 Code | | 53.6 | | | Kimi K2.7 Code (Kimi Code CLI) | 2026-06-11 |
| ProgramBench Leaderboard | | 52.1 | 0.0 | 2.5 | Claude Opus 4.6 (mini-SWE-agent) | 2026-02-05 |
| GLM-5.1: Towards Long-Horizon Tasks | ✓ Link | 50.9 | | | GLM-5.1 (Max) (Claude Code) | 2026-04-03 |
| ProgramBench Leaderboard | | 50.9 | 0.0 | 3.0 | Claude Opus 4.7 (mini-SWE-agent) | 2026-04-16 |
| Seed2.1 Model Card: Agentic Intelligence for Productivity | | 50.3 | | | Seed2.1 Pro | 2026-06-23 |
| MindForge: Teaching Small Language Models Whole-Life-Cycle Software Engineering via Source-Free Program Synthesis | | 49.51 | | | MindForge-27B (Mini-SWE-Agent) | 2026-07-29 |
| ProgramBench Leaderboard | | 47.5 | 0.0 | 1.0 | Claude Sonnet 4.6 (mini-SWE-agent) | 2026-02-17 |
| ProgramBench Leaderboard | ✓ Link | 47.0 | 0.0 | 4.0 | Muse Spark 1.1 (xhigh) (mini-SWE-agent) | 2026-07-09 |
| Gemini 3.1 Pro: A smarter model for your most complex tasks | | 39.5 | | | Gemini 3.1 Pro (Max) (Claude Code) | 2026-02-19 |
| ProgramBench Leaderboard | | 37.7 | 0.0 | 0.0 | GPT 5.4 (mini-SWE-agent) | 2026-03-05 |
| ProgramBench Leaderboard | | 31.8 | 0.0 | 0.0 | Gemini 3 Flash (mini-SWE-agent) | 2025-11-18 |
| ProgramBench Leaderboard | | 30.0 | 0.0 | 0.0 | Claude Haiku 4.5 (mini-SWE-agent) | 2025-10-15 |
| ProgramBench Leaderboard | | 16.4 | 0.0 | 0.0 | GPT 5.4 mini (mini-SWE-agent) | 2026-03-17 |
| ProgramBench Leaderboard | | 16.0 | 0.0 | 0.0 | GPT 5 mini (mini-SWE-agent) | 2025-08-07 |
| GLM-5.3: Frontier Coding with Emergent Cyber Capabilities | ✓ Link | | | 19.0 | GLM-5.3 | 2026-08-14 |
| Hy4 preview | ✓ Link | | | 17.5 | Hy4 preview (Claude Code) | 2026-08-28 |
| DeepSeek-V4.1-Flash: Pushing the Limits of KV Cache Compression | ✓ Link | | | 20.3 | DeepSeek-V4.1-Flash Max (DeepSeek Harness (Minimal)) | 2026-09-17 |
| Muse Wonder: A Frontier Model For Personal & Professional AI Agents | | | | 59.0 | Muse Wonder Full (max) (mini-SWE-agent) | 2026-10-12 |
| Muse Wonder: A Frontier Model For Personal & Professional AI Agents | | | | 50.0 | Muse Wonder (max) (mini-SWE-agent) | 2026-10-12 |
| DeepSeek-V4.1-Flash: Pushing the Limits of KV Cache Compression | ✓ Link | | | 29.1 | DeepSeek-V4.1-Flash-1017 | 2026-09-17 |