| Introducing Claude Opus 5 | ✓ Link | 96.0 | Claude Opus 5 | 2026-07-24 |
| Claude Fable 5 and Claude Mythos 5 | ✓ Link | 95.5 | Claude Mythos 5 | 2026-06-09 |
| Introducing Ember-1 | | 92.2 | Ember-1 | 2026-09-23 |
| Introducing Claude Opus 4.8 | | 88.6 | Claude Opus 4.8 | 2026-05-28 |
| Introducing Claude Opus 4.7 | | 87.6 | Claude Opus 4.7 | 2026-04-16 |
| Macaron-V1: Towards Open Continual Learning with Self-Improvement and Mixture-of-LoRA | ✓ Link | 85.6 | Macaron-V1-Venti (Claude Code) | 2026-08-10 |
| Lean4AgenticRL: Advancing Agentic RL with Verifiable Multi-level Process Rewards | ✓ Link | 84.2 | LAR-27B (Claude-code) | 2026-10-15 |
| Ornith-1.0: Self-Scaffolding LLMs for Agentic Coding | ✓ Link | 82.4 | Ornith-1.0-397B (OpenHands) | 2026-06-21 |
| NVIDIA-labs OO Agents: Native Python Object-Oriented Agents | | 82.2 | NOOA agent (GPT-5.5, xhigh) | 2026-07-22 |
| Introducing Claude Opus 4.5 | | 80.9 | Claude Opus 4.5 | 2025-11-24 |
| Introducing Beam: Reflection's 501B open-weight model | | 80.9 | Beam | 2026-10-05 |
| Introducing Claude Opus 4.6 | | 80.84 | Claude Opus 4.6 | 2026-02-05 |
| Gemini 3.1 Pro: A smarter model for your most complex tasks | | 80.6 | Gemini 3.1 Pro | 2026-02-19 |
| DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence | ✓ Link | 80.6 | DeepSeek-V4-Pro Max | 2026-04-26 |
| MiniMax M3: Frontier Coding, 1M Context, Native Multimodality — All in One Model | ✓ Link | 80.5 | MiniMax M3 (Claude Code) | 2026-05-31 |
| MiniMax-M2.5 | ✓ Link | 80.2 | MiniMax-M2.5 (Claude Code) | 2026-02-12 |
| Kimi K2.6 | ✓ Link | 80.2 | Kimi K2.6 (In-house SWE-agent-adapted framework) | 2026-04-20 |
| Inkling-Small | ✓ Link | 80.2 | Inkling-Small (Bash-only) | 2026-07-27 |
| Introducing GPT-5.2 | | 80.0 | GPT-5.2 | 2025-12-11 |
| Introducing Claude Sonnet 4.6 | | 79.6 | Claude Sonnet 4.6 | 2026-02-17 |
| KAT-Coder-V2 Technical Report | | 79.6 | KAT-Coder-V2 (Claude Code) | 2026-03-29 |
| DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence | ✓ Link | 79.0 | DeepSeek-V4-Flash Max | 2026-04-26 |
| Ornith-1.5: From Self-Scaffolding to Self-Improvement | | 79.0 | Ornith-1.5-35B-A3B (OpenHands) | 2026-08-20 |
| MiMo-V2.5-Pro | ✓ Link | 78.9 | MiMo-V2.5-Pro | 2026-04-27 |
| Qwen3.6-Plus: Towards Real World Agents | | 78.8 | Qwen3.6-Plus | 2026-04-02 |
| dots3-note Preview: A Small but Mighty Step Toward Long-Horizon Agency in Real Life | ✓ Link | 78.4 | dots3-note Preview (live-swe-agent) | 2026-08-13 |
| Hy3 | ✓ Link | 78.0 | Hy3 (SWE-agent) | 2026-07-02 |
| GLM-5: From Vibe Coding to Agentic Engineering | ✓ Link | 77.8 | GLM-5 (OpenHands) | 2026-02-11 |
| Inkling: Our open-weights model | ✓ Link | 77.6 | Inkling (bash-only harness) | 2026-07-15 |
| Marathoner: Ultra-Long-Horizon Autonomous Intelligence | | 77.5 | Marathoner-9B (Claude Code) | 2026-09-28 |
| Introducing Claude Sonnet 4.5 | | 77.2 | Claude Sonnet 4.5 | 2025-09-29 |
| Qwen3.6 | ✓ Link | 77.2 | Qwen3.6-27B (Internal agent scaffold) | 2026-04-21 |
| Kimi K2.5: Visual Agentic Intelligence | ✓ Link | 76.8 | Kimi K2.5 (Internal agent scaffold) | 2026-02-02 |
| ModularRSI: Modular and Generalizable Recursive Harness Self-Improvement | ✓ Link | 76.45 | ModularRSI (Harbor) | 2026-09-14 |
| Qwen3.5: Towards Native Multimodal Agents | ✓ Link | 76.4 | Qwen3.5-397B-A17B (Internal agent scaffold) | 2026-02-16 |
| A new era of intelligence with Gemini 3 | | 76.2 | Gemini 3 Pro (single attempt) | 2025-11-18 |
| Motif 3: Technical Report | ✓ Link | 76.2 | Motif 3 (mini-SWE-agent) | 2026-08-10 |
| Muse Glimmer 30B: An On-Device Agentic Model | ✓ Link | 76.0 | Muse Glimmer-30B (High) | 2026-10-17 |
| Ornith-1.0: Self-Scaffolding LLMs for Agentic Coding | ✓ Link | 75.6 | Ornith-1.0-35B (OpenHands) | 2026-06-21 |
| Introducing GPT-5 | | 74.9 | GPT-5 (High) | 2025-08-07 |
| InCoder-32B: Code Foundation Model for Industrial Scenarios | ✓ Link | 74.8 | InCoder-32B (32B) | 2026-03-17 |
| Laguna M.1/XS.2 Technical Report | ✓ Link | 74.6 | Laguna M.1 (pool) | 2026-05-26 |
| Claude Opus 4.1 | | 74.5 | Claude Opus 4.1 | 2025-08-05 |
| Step 3.5 Flash: Open Frontier-Level Intelligence with 11B Active Parameters | ✓ Link | 74.4 | Step-3.5-Flash | 2026-02-11 |
| Hy3 preview | ✓ Link | 74.4 | Hy3 preview | 2026-04-23 |
| Ling and Ring 2.6 Technical Report: Efficient and Instant Agentic Intelligence at Trillion-Parameter Scale | ✓ Link | 74.00 | Ring-2.6-1T (High) (Claude Code) | 2026-06-13 |
| MindForge: Teaching Small Language Models Whole-Life-Cycle Software Engineering via Source-Free Program Synthesis | | 73.84 | MindForge-27B (Mini-SWE-Agent) | 2026-07-29 |
| MAI-Thinking-1: Building a Hill-Climbing Machine | | 73.5 | MAI-Thinking-1 | 2026-06-02 |
| MiMo-V2-Flash Technical Report | ✓ Link | 73.4 | MiMo-V2-Flash (Thinking) | 2026-01-06 |
| Qwen3.6 | ✓ Link | 73.4 | Qwen3.6-35B-A3B (Internal agent scaffold) | 2026-04-21 |
| DeepSeek-V3.2 Technical Report | ✓ Link | 73.1 | DeepSeek-V3.2 (Internal framework) | 2025-12-02 |
| Introducing Claude 4 | | 72.7 | Claude Sonnet 4 | 2025-05-22 |
| Introducing Claude 4 | | 72.5 | Claude Opus 4 | 2025-05-22 |
| Qwen3.5: Towards Native Multimodal Agents | ✓ Link | 72.4 | Qwen3.5-27B | 2026-02-16 |
| Ling-2.6-1T: A Trillion-Parameter Comprehensive Flagship Model for Complex Tasks | | 72.20 | Ling-2.6-1T (OpenHands) | 2026-04-29 |
| SkillForge: Self-Distilling Agents for Project-Specific Issue Resolution | ✓ Link | 72.2 | SkillForge (DeepSeek-V3.2) (Mini-SWE-Agent) (extra data) | 2026-08-19 |
| Bilevel Coordinated Reflection: A Game-Theoretic Approach to Multi-Agent LLM Systems | ✓ Link | 72.2 | Bilevel SRMA (Kimi K2.5) (official test harness) | 2026-09-02 |
| Qwen3.5: Towards Native Multimodal Agents | ✓ Link | 72.0 | Qwen3.5-122B-A10B | 2026-02-16 |
| Nemotron 3 Ultra: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Model for Agentic Reasoning | ✓ Link | 71.9 | Nemotron 3 Ultra (OpenHands) | 2026-06-03 |
| Qwen3-Coder-Next Technical Report | ✓ Link | 71.3 | Qwen3-Coder-Next (OpenHands) | 2026-02-28 |
| Priming Base Models for Agentic Post-Training: An Open Midtraining Recipe | | 71.1 | NVIDIA Nemotron 3.5 Super VL 120B-A12B BF16 | 2026-10-15 |
| Introducing Laguna XS 2.1 | ✓ Link | 70.9 | Laguna XS 2.1 (pool) | 2026-07-02 |
| PreciseCoder: Faithful and Generalizable Debugging Beyond Unit-Test Rewards | | 70.8 | PreciseCoder-27B (SWE-agent) | 2026-10-13 |
| Mach-Mind-4-Flash Technical Report | | 70.60 | Mach-Mind-4-Flash | 2026-07-10 |
| Introducing K2 Horizon: Frontier Performance, Radically Open | ✓ Link | 70.6 | K2-Horizon-7B | 2026-09-03 |
| InCoder-32B-Thinking: Industrial Code World Model for Thinking | ✓ Link | 70.4 | InCoder-32B-Thinking | 2026-04-03 |
| Solar Open 2 Technical Report | ✓ Link | 70.4 | Solar Open 2 250B-A15B | 2026-07-22 |
| LEGO-RL: Harness-Native Reinforcement Learning for Coding Agents | ✓ Link | 70.4 | Lego-RL-Qwen3.5-35B-A3B (OpenHands SDK) (extra data) | 2026-08-18 |
| Laguna M.1/XS.2 Technical Report | ✓ Link | 69.9 | Laguna XS.2 (pool) | 2026-05-26 |
| MemGovern: Enhancing Code Agents through Learning from Governed Human Experiences | ✓ Link | 69.8 | MemGovern (Claude-4-Sonnet) (SWE-Agent) | 2026-01-11 |
| Self-Evolving Coding Agents via Harness-Policy Optimization | ✓ Link | 69.5 | Self-HPO-Qwen3.5-35B-A3B | 2026-10-16 |
| Ornith-1.0: Self-Scaffolding LLMs for Agentic Coding | ✓ Link | 69.4 | Ornith-1.0-9B (OpenHands) | 2026-06-21 |
| Qwen3.5: Towards Native Multimodal Agents | ✓ Link | 69.2 | Qwen3.5-35B-A3B | 2026-02-16 |
| UI-TARS-2 Technical Report: Advancing GUI Agent with Multi-Turn Reinforcement Learning | ✓ Link | 68.7 | UI-TARS-2 (w/ GUI-SDK) | 2025-09-02 |
| ROSS: Relearning from Self-Generated Rollouts through Selective Supervision | | 68.4 | ROSS (Qwen3.6-35B-A3B) (Codex CLI) | 2026-09-28 |
| Qwen-AgentWorld: Language World Models for General Agents | ✓ Link | 67.86 | w/ LWM RL on Qwen3.5-35B-A3B-SFT (Internal agent scaffold (bash and file-edit tools)) | 2026-06-23 |
| Introducing North Mini Code: Cohere's first model for developers | ✓ Link | 67.6 | North Mini Code (SWE-agent) | 2026-06-09 |
| Orchard: An Open-Source Agentic Modeling Framework | ✓ Link | 67.5 | Qwen3-30B-A3B-Thinking (Orchard-SWE, SFT+RL) (mini-swe-agent) | 2026-05-14 |
| PACT: From Credit Assignment to Critic Alignment | ✓ Link | 67.4 | PACT (Qwen3.6-35B-A3B) (Codex through Harbor) | 2026-09-22 |
| Kolibri: A Sovereign European Model on the Pareto Frontier | ✓ Link | 66.4 | Kolibri (OpenCode) | 2026-10-03 |
| daVinci-Env: Open SWE Environment Synthesis at Scale | ✓ Link | 66.0 | OpenSWE-72B (SWE-Agent) | 2026-03-13 |
| PivotOPD: Learning to Recover from Pivotal Mistakes in Multi-Turn Agents | | 66.0 | PivotOPD (SWE-Bench Verified) | 2026-09-30 |
| Rufus-Air: An Open LLM Post-Training Recipe | | 65.6 | Rufus-Air | 2026-09-24 |
| LoopCoder-v2: Only Loop Once for Efficient Test-Time Computation Scaling | ✓ Link | 64.4 | LoopCoder-v2 | 2026-06-16 |
| GLM-4.5: Agentic, Reasoning, and Coding (ARC) Foundation Models | ✓ Link | 64.2 | GLM-4.5 (355B) (OpenHands) | 2025-08-08 |
| Nanbeige4.2-3B: Unlocking Agentic Capabilities in a Compact Model | ✓ Link | 63.6 | Nanbeige4.2-3B (OpenHands) | 2026-07-21 |
| gpt-oss-120b & gpt-oss-20b Model Card | | 62.4 | gpt-oss-120b (High) | 2025-08-08 |
| daVinci-Env: Open SWE Environment Synthesis at Scale | ✓ Link | 62.4 | OpenSWE-32B (SWE-Agent) | 2026-03-13 |
| Claude 3.7 Sonnet and Claude Code | | 62.3 | Claude 3.7 Sonnet | 2025-02-24 |
| Kwai Keye-VL-2.0 Technical Report | ✓ Link | 62.0 | Keye-VL-2.0-30B-A3B | 2026-06-09 |
| Dockerless: Environment-Free Program Verifier for Coding Agents | | 62.0 | Dockerless-RL-9B (OpenHands) | 2026-06-26 |
| Closing the Loop: Practical Training Recipes for Looped Language Models | | 61.8 | Dory1.0-4B (NeMo Gym / Nemo Evaluator SDK) | 2026-09-30 |
| FrogNano: Training a 4B Coding Agent via Online Task Synthesis | ✓ Link | 61.5 | FrogNano-4B-2609 (Leaf) | 2026-09-07 |
| SWE-Master: Unleashing the Potential of Software Engineering Agents via Post-Training | ✓ Link | 61.4 | SWE-Master-32B-RL (R2E-Gym) | 2026-02-03 |
| Ling and Ring 2.6 Technical Report: Efficient and Instant Agentic Intelligence at Trillion-Parameter Scale | ✓ Link | 61.20 | Ling-2.6-flash (Claude Code) | 2026-06-13 |
| NVIDIA Nemotron 3: Efficient and Open Intelligence | ✓ Link | 60.47 | Nemotron 3 Super (OpenHands) | 2025-12-24 |
| LongCat-Flash Technical Report | ✓ Link | 60.40 | LongCat-Flash (OpenHands) | 2025-09-01 |
| SWE-Lego: Pushing the Limits of Supervised Fine-tuning for Software Issue Resolving | ✓ Link | 60.4 | SWE-Lego-Qwen3-32B (OpenHands) | 2026-01-04 |
| NVIDIA Nemotron 3.5 Lightning | ✓ Link | 60.0 | NVIDIA Nemotron 3.5 Lightning 30B-A3B (OpenCode v1.17.8) | 2026-08-11 |
| Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities | | 59.6 | Gemini 2.5 Pro (single attempt) | 2025-07-07 |
| ReTerm: Recovery Training for Self-Improving Terminal Agents | ✓ Link | 58.4 | ReTerm-9B (ReAct) | 2026-10-13 |
| GLM-4.5: Agentic, Reasoning, and Coding (ARC) Foundation Models | ✓ Link | 57.6 | GLM-4.5-Air (OpenHands) | 2025-08-08 |
| Let It Flow: Agentic Crafting on Rock and Roll, Building the ROME Model within an Open Agentic Learning Ecosystem | ✓ Link | 57.4 | ROME (30B MoE, activated 3B) (iFlow CLI) | 2025-12-31 |
| Granite 4.2 LLMs: How They're Built | ✓ Link | 57 | Granite-4.2-30B | 2026-08-25 |
| Agent Lightning v1.0: Towards Harnessed Agentic RL | ✓ Link | 56.4 | Agent Lightning v1.0 (Qwen3.5-9B) (mini-SWE-agent) (extra data) | 2026-08-18 |
| SERA: Soft-Verified Efficient Repository Agents | ✓ Link | 54.2 | SERA-32B (SWE-agent) | 2026-01-28 |
| OpenThoughts-Agent: Data Recipes for Agentic Models | ✓ Link | 54.0 | OpenThinkerAgent-32B (Terminus-2) | 2026-06-23 |
| Tmax: A simple recipe for terminal agents | ✓ Link | 53.5 | Tmax 9B (Vanillux2Agent) | 2026-06-22 |
| LFM2.5-24B-A2B: Built for Fast On-Device Agents | | 51.00 | LFM2.5-24B-A2B (OpenCode) | 2026-10-13 |
| K-EXAONE Technical Report | ✓ Link | 49.4 | K-EXAONE (236B, reasoning) (mini-SWE-agent) | 2026-01-05 |
| DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning | ✓ Link | 49.2 | DeepSeek-R1 | 2025-01-22 |
| Introducing OpenAI o1 | | 48.9 | OpenAI o1 (High) | 2024-09-12 |
| Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities | | 48.9 | Gemini 2.5 Flash (single attempt) | 2025-07-07 |
| Unified Audio Intelligence Without Regressing on Text Intelligence | | 48.2 | Audex 30B-A3B (OpenHands) | 2026-07-06 |
| MiniCPM5-2B | ✓ Link | 46.4 | MiniCPM5-2B | 2026-09-07 |
| AEM: Adaptive Entropy Modulation for Multi-Turn Agentic Reinforcement Learning | | 43.7 | DeepSWE + AEM (Qwen3-32B) | 2026-05-08 |
| DeepSeek-V3 Technical Report | ✓ Link | 42.0 | DeepSeek-V3 (Agentless) | 2024-12-27 |
| FC-SWE: Failure-Conditioned RL for Long-Horizon Software Engineering Agents | | 41.7 | FC-SWE (Qwen3.5-4B) (SWE-agent) | 2026-10-06 |
| Polar: Agentic RL on Any Harness at Scale | ✓ Link | 40.4 | Polar RL (Qwen3.5-4B) (Pi) | 2026-05-22 |
| SWE-smith: Scaling Data for Software Engineering Agents | ✓ Link | 40.2 | SWE-agent-LM-32B | 2025-04-30 |
| Function-Aware Fill-in-the-Middle as Mid-Training for Coding Agent Foundation Models | ✓ Link | 35.0 | Qwen3-8B + FIM-Midtrain + SWE-Lego (OpenHands) | 2026-07-14 |
| PivotRL: High Accuracy Agentic Post-Training at Low Compute Cost | | 32.67 | Qwen3-30B-A3B-Thinking-2507 + PivotRL (OpenHands) | 2026-03-22 |
| SERA: Soft-Verified Efficient Repository Agents | ✓ Link | 31.7 | SERA-8B (SWE-agent) | 2026-01-28 |
| Context-Aware RL for Agentic and Multimodal LLMs | ✓ Link | 30.2 | ContextRL (trained from Klear-AgentForge-8B) (mini-SWE-agent) | 2026-06-15 |
| SWE-AGILE: A Software Agent Framework for Efficiently Managing Dynamic Reasoning Context | ✓ Link | 30.06 | SWE-AGILE (SFT) (Qwen3-14B) (R2EGym) | 2026-04-13 |
| AgentDevel: Reframing Self-Evolving LLM Agents as Release Engineering | | 30.0 | AgentDevel (final) (Claude Code) | 2026-01-08 |
| Revisiting DAgger in the Era of LLM-Agents | | 29.8 | Qwen3-8B + DAgger-style Training (OpenHands) | 2026-05-13 |
| Single-Rollout Asynchronous Optimization for Agentic Reinforcement Learning | | 29.8 | Qwen3-30B-A3B + SAO (OpenHands) | 2026-07-08 |
| ProRL Agent: Rollout-as-a-Service for RL Training of Multi-Turn LLM Agents | ✓ Link | 23.6 | ProRL Agent-14B (RL) | 2026-03-19 |
| What Should Agents Say? Action-state Communication for Efficient Multi-Agent Systems | ✓ Link | 23.0 | PACT (Qwen3-14B) — OpenHands | 2026-06-03 |
| Youtu-LLM: Unlocking the Native Agentic Potential for Lightweight Large Language Models | ✓ Link | 17.7 | Youtu-LLM 2B | 2025-12-31 |