| Atria Dawn: The Dawn of Agentic Superintelligence | ✓ Link | 92.5 | Atria Dawn Preview | 2026-09-14 |
| GPT-6 Astra: A new generation of intelligence | ✓ Link | 91.5 | GPT-6 Astra | 2026-09-03 |
| Kimi K3: Open Frontier Intelligence | ✓ Link | 91.2 | Kimi K3 Max | 2026-07-27 |
| Introducing GPT-5.5 | | 90.1 | GPT-5.5 Pro | 2026-04-23 |
| NexForge: Scaling Agent Capabilities through Requirement-Driven Task Synthesis for LLMs | ✓ Link | 89.7 | Nex-N2.5-Pro | 2026-07-17 |
| Introducing GPT-5.4 | | 89.3 | GPT-5.4 Pro | 2026-03-05 |
| REMORY: Learning Residual Memory for Context Compaction | ✓ Link | 89.0 | GLM-5.3-Flash + REMORY | 2026-10-08 |
| MiroThinker-1.7 & H1: Towards Heavy-Duty Research Agents via Verification | | 88.2 | MiroThinker-H1 | 2026-03-16 |
| Gemini 3.1 Pro: A smarter model for your most complex tasks | | 85.9 | Gemini 3.1 Pro (w/ tools) | 2026-02-19 |
| Hy3 | ✓ Link | 84.2 | Hy3 | 2026-07-02 |
| AREX-2: Advancing Self-Improving Agents through Long-Horizon Reflective Tasks | ✓ Link | 84.0 | AREX-2 | 2026-09-29 |
| MiniMax M3: Frontier Coding, 1M Context, Native Multimodality — All in One Model | ✓ Link | 83.52 | MiniMax M3 | 2026-05-31 |
| DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence | ✓ Link | 83.4 | DeepSeek-V4-Pro Max | 2026-04-26 |
| dots3-note Preview: A Small but Mighty Step Toward Long-Horizon Agency in Real Life | ✓ Link | 83.3 | dots3-note Preview (w/ context management) | 2026-08-13 |
| Kimi K2.6 | ✓ Link | 83.2 | Kimi K2.6 | 2026-04-20 |
| AREX: Towards a Recursively Self-Improving Agent for Deep Research | ✓ Link | 82.5 | AREX-Base (122B-A10B) | 2026-07-23 |
| Sage-A1.5: From Long-Horizon Execution to Scientific Exploration with Capability Composition | ✓ Link | 81.7 | Sage-A1.5 | 2026-10-20 |
| Introducing LongCat-2.0 | ✓ Link | 79.9 | LongCat-2.0 | 2026-07-05 |
| GLM-5.1: Towards Long-Horizon Tasks | ✓ Link | 79.3 | GLM-5.1 (w/ context manage) | 2026-04-03 |
| Introducing Claude Opus 4.7 | | 79.3 | Claude Opus 4.7 | 2026-04-16 |
| Introducing GPT-5.2 | | 77.9 | GPT-5.2 Pro | 2025-12-11 |
| The MiniMax-M2 Series: Mini Activations Unleashing Max Real-World Intelligence | ✓ Link | 77.8 | MiniMax-M2.7 | 2026-05-26 |
| Inkling-Small | ✓ Link | 77.4 | Inkling-Small (with context management) | 2026-07-27 |
| Introducing Beam: Reflection's 501B open-weight model | | 77.4 | Beam (w/ context management) | 2026-10-05 |
| Seed2.0 Model Card: Towards Intelligence Frontier for Real-World Complexity | | 77.3 | Seed2.0 Pro | 2026-06-30 |
| Inkling: Our open-weights model | ✓ Link | 77.1 | Inkling | 2026-07-15 |
| MiniMax-M2.5 | ✓ Link | 76.3 | MiniMax-M2.5 | 2026-02-12 |
| GLM-5: From Vibe Coding to Agentic Engineering | ✓ Link | 75.9 | GLM-5 (w/ context manage) | 2026-02-11 |
| Step 3.7 Flash | ✓ Link | 75.82 | Step 3.7 Flash | 2026-05-29 |
| Scaling the Horizon, Not the Parameters: Reaching Trillion-Parameter Performance with a 35B Agent | ✓ Link | 75.5 | Agents-A1 | 2026-06-29 |
| Kimi K2.5: Visual Agentic Intelligence | ✓ Link | 74.9 | Kimi K2.5 (w/ ctx manage) | 2026-02-02 |
| EvoMaster: A Foundational Agent Framework for Building Evolving Autonomous Scientific Agents at Scale | ✓ Link | 73.3 | EvoMaster | 2026-04-19 |
| DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence | ✓ Link | 73.2 | DeepSeek-V4-Flash Max | 2026-04-26 |
| Introducing K2 Horizon: Frontier Performance, Radically Open | ✓ Link | 72.8 | K2-Horizon-375B-A23B | 2026-09-03 |
| FORT-Searcher: Synthesizing Shortcut-Resistant Search Tasks for Training Deep Search Agents | ✓ Link | 72.2 | FORT-Searcher | 2026-06-10 |
| Agentic Aggregation for Parallel Scaling of Long-Horizon Agentic Tasks | ✓ Link | 71.33 | AggAgent (MiniMax-M2.5) | 2026-04-13 |
| AgentFugue: Agent Scaling for Long-Horizon Tasks through Collective Reasoning | | 71.2 | AgentFugue (DeepSeek-v4-Flash) | 2026-05-23 |
| Step 3.5 Flash: Open Frontier-Level Intelligence with 11B Active Parameters | ✓ Link | 69.0 | Step-3.5-Flash (w/ context manager) | 2026-02-11 |
| Qwen3.5: Towards Native Multimodal Agents | ✓ Link | 69.0 | Qwen3.5-397B-A17B | 2026-02-16 |
| SearchSwarm: Towards Delegation Intelligence in Agentic LLMs for Long-Horizon Deep Research | ✓ Link | 68.1 | SearchSwarm-30B-A3B | 2026-06-08 |
| GLM-5.1: Towards Long-Horizon Tasks | ✓ Link | 68.0 | GLM-5.1 | 2026-04-03 |
| DeepSeek-V3.2 Technical Report | ✓ Link | 67.6 | DeepSeek-V3.2 (w/ context management) | 2025-12-02 |
| Ornith-1.5: From Self-Scaffolding to Self-Improvement | | 67.6 | Ornith-1.5-35B-A3B | 2026-08-20 |
| Hy3 preview | ✓ Link | 67.1 | Hy3 preview | 2026-04-23 |
| QUEST: Training Frontier Deep Research Agents with Fully Synthetic Tasks | ✓ Link | 64.6 | QUEST-35B | 2026-05-22 |
| Qwen3.5: Towards Native Multimodal Agents | ✓ Link | 63.8 | Qwen3.5-122B-A10B | 2026-02-16 |
| AgentSwing: Adaptive Parallel Context Management Routing for Long-Horizon Web Agents | | 62.5 | AgentSwing (DeepSeek-v3.2) | 2026-03-29 |
| Mach-Mind-4-Flash Technical Report | | 61.70 | Mach-Mind-4-Flash | 2026-07-10 |
| Qwen3.5: Towards Native Multimodal Agents | ✓ Link | 61.0 | Qwen3.5-27B | 2026-02-16 |
| Kimi K2.5: Visual Agentic Intelligence | ✓ Link | 60.6 | Kimi K2.5 | 2026-02-02 |
| SAM: State-Adaptive Memory for Long-Horizon Reasoning Agent | ✓ Link | 56.5 | GLM-4.7 + SAM | 2026-05-23 |
| HarnessDev: Can LLMs Create and Evolve Their Own Agent Harness? | | 55.4 | HarnessDev Creation (GPT-5.5 creator, Gemini 3.1 Pro executor) | 2026-09-01 |
| ABSeeker: Training Long-Horizon Search Agents via Answer-Backtracked Credit Assignment | | 55.3 | ABSeeker (Qwen3.5-4B, context management) | 2026-08-05 |
| RE-TRAC: REcursive TRAjectory Compression for Deep Search Agents | ✓ Link | 53.0 | RE-TRAC-30B-A3B | 2026-02-02 |
| Step 3.5 Flash: Open Frontier-Level Intelligence with 11B Active Parameters | ✓ Link | 51.6 | Step-3.5-Flash | 2026-02-11 |
| Introducing OpenAI o3 and o4-mini | ✓ Link | 51.5 | OpenAI o4-mini (w/ Python) | 2025-04-16 |
| Search More, Think Less: Rethinking Long-Horizon Agentic Search for Efficiency and Generalization | ✓ Link | 48.6 | SMTL-300 (30B) | 2026-02-26 |
| Agent-Editing World Model: Rethinking World Modeling for LLM Agents | ✓ Link | 48.1 | EditAct (Qwen3.5-35B-A3B) | 2026-09-23 |
| OpenSeeker-v2: Pushing the Limits of Search Agents with Informative and High-Difficulty Trajectories | ✓ Link | 46.0 | OpenSeeker-v2-30B-SFT | 2026-05-05 |
| MiMo-V2-Flash Technical Report | ✓ Link | 45.4 | MiMo-V2-Flash (Thinking) | 2026-01-06 |
| Nemotron 3 Ultra: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Model for Agentic Reasoning | ✓ Link | 44.4 | Nemotron 3 Ultra | 2026-06-03 |
| Mind DeepResearch Technical Report | | 42.8 | MindDR-v1.5-30B-A3B | 2026-04-17 |
| Mosaic: Multi-Session Agentic Reinforcement Learning | ✓ Link | 38.86 | Mosaic (Qwen3.5-4B) | 2026-10-15 |
| NVIDIA Nemotron 3.5 Lightning | ✓ Link | 36.97 | NVIDIA Nemotron 3.5 Lightning 30B-A3B | 2026-08-11 |
| SlimSearcher: Training Efficiency-Aware Web Agents via Adaptive Reward Gating | ✓ Link | 32.8000 | SlimSearcher (SFT+RL) | 2026-06-05 |
| Nested Browser-Use Learning for Agentic Information Seeking | | 31.6 | NestBrowse-30B-A3B | 2025-12-29 |
| K-EXAONE Technical Report | ✓ Link | 31.4 | K-EXAONE (236B, reasoning) | 2026-01-05 |
| Marco DeepResearch: Unlocking Efficient Deep Research Agents via Verification-Centric Design | ✓ Link | 31.4 | Marco-DeepResearch-8B | 2026-03-30 |
| NVIDIA Nemotron 3: Efficient and Open Intelligence | ✓ Link | 31.28 | Nemotron 3 Super | 2025-12-24 |
| IterSynth: Rethinking Deep Search Agents via Role-Decoupled Iterative Synthesis | ✓ Link | 30.9 | IterSynth-8B | 2026-09-24 |
| UI-TARS-2 Technical Report: Advancing GUI Agent with Multi-Turn Reinforcement Learning | ✓ Link | 29.6 | UI-TARS-2 (w/ GUI-SDK) | 2025-09-02 |
| OpenSeeker: Democratizing Frontier Search Agents by Fully Open-Sourcing Training Data | ✓ Link | 29.5 | OpenSeeker-v1-30B-SFT | 2026-03-16 |
| Kolibri: A Sovereign European Model on the Pareto Frontier | ✓ Link | 29.4 | Kolibri | 2026-10-03 |
| GLM-4.5: Agentic, Reasoning, and Coding (ARC) Foundation Models | ✓ Link | 26.4 | GLM-4.5 (355B) | 2025-08-08 |
| OpenResearcher: A Fully Open Pipeline for Long-Horizon Deep Research Trajectory Synthesis | ✓ Link | 26.3 | OpenResearcher-30B-A3B | 2026-03-17 |
| Efficient Agentic Reasoning Through Self-Regulated Simulative Planning | ✓ Link | 22.5 | SR 2 AM-v1.0-30B | 2026-05-21 |
| GLM-4.5: Agentic, Reasoning, and Coding (ARC) Foundation Models | ✓ Link | 21.3 | GLM-4.5-Air | 2025-08-08 |
| Nanbeige4.1-3B: A Small General Model that Reasons, Aligns, and Acts | ✓ Link | 19.12 | Nanbeige4.1-3B (w/ tools) | 2026-02-13 |
| PivotRL: High Accuracy Agentic Post-Training at Low Compute Cost | | 11.3 | Qwen3-30B-A3B-Thinking-2507 + PivotRL | 2026-03-22 |
| Inference-Time Scaling of Verification: Self-Evolving Deep Research Agents via Test-Time Rubric-Guided Verification | ✓ Link | 10.0 | DeepVerifier (with Claude-3.7-Sonnet backbone) | 2026-01-22 |
| SAGE: Steerable Agentic Data Generation for Deep Search with Execution Feedback | ✓ Link | 2.6 | QWEN-7B (trained on SAGE) | 2026-01-26 |