Understand this story
On November 2, 2026, the Meituan LongCat team released LongCat-2.5, a native multimodal model for long-horizon agentic tasks, with a total parameter count of 1.6 trillion and approximately 48 billion parameters activated per token. According to a Phoenix Technology report, the model leads Claude-Opus-5 and GPT-5.6-Sol on benchmarks published by Meituan such as AutomationBench and LongCatClawBench, and scores 38.5 on Terminal Bench v4.0, surpassing GPT-5.6-Sol but trailing Claude-Opus-5. The model converts visual signals into multi-level discrete tokens and processes them uniformly with text, and redesigns its parallelism strategy for domestic accelerators with less than 80GB of VRAM per card, increasing single-card throughput for long-sequence SFT by 86% over the DDP baseline. China.com Technology reported the same day that the model was trained entirely on domestic compute, that its evaluation results fall between Claude-Opus-4.8 and Claude-Opus-5, and that it surpasses Claude-Opus-5 on some evaluations; Meituan also launched "Miaoshou," an AI business operations workbench based on the LongCat series.
AI-generated from public reports
Latest development · Report
Meituan Releases LongCat-2.5 Large Model, 1.6 Trillion Parameters, Trained Entirely on Domestic ComputeDevelopments
1 confirmed developments
Meituan LongCat releases native multimodal model LongCat-2.5
Source:Phoenix Technology
Reports and discussion timeline
Browse reports and discussion by time. Their evidence levels remain distinct.
Nov 3
Caixin
Meituan Releases LongCat-2.5 Large Model, 1.6 Trillion Parameters, Trained Entirely on Domestic Compute
On November 2, Meituan released its new-generation large model LongCat-2.5, with 1.6 trillion total parameters and approximately 48 billion parameters activated per token, continuing the feature of LongCat-2.0 in being trained entirely on domestic compute. Meituan says the model is a native multimodal model for long-horizon agent tasks, strengthening native visual understanding and agentic coding capabilities on the basis of LongCat-2.0, and supporting large-scale long-horizon RL by optimizing long-context training on domestic compute along with W4A4 inference, training-inference numerical consistency, and long-context rollout optimization. Meituan says its results on multiple mainstream benchmark evaluations fall between those of Opus 4.8 and Opus 5, released by Anthropic in May and July of this year.
Phoenix Technology
Meituan Releases LongCat-2.5 Large Model, Topping Multiple Agent Benchmarks
Meituan has launched its new-generation large model LongCat-2.5, which ranks first on multiple of 9 evaluation benchmarks across three categories: coding, general tasks, and visual Agent. The model is a native multimodal model designed for long-horizon agent tasks, with 1.6 trillion total parameters and approximately 48 billion parameters activated per token. Building on LongCat-2.0, it strengthens native visual understanding and agentic coding capabilities. It scored 51.8 on AutomationBench, surpassing Claude-Opus-5's 50.3, and also ranked first on BrowseComp with a score of 91.9; on 3DCodeBench it scored 55.4 on text-to-3D. It still lags behind leading overseas models on coding and terminal instruction tasks.
Testing Catalog
Meituan releases the LongCat-2.5 multimodal model, with 1.6T total parameters and 49B active
Meituan has released LongCat-2.5, a natively multimodal agentic model with 1.6T total parameters and about 48B active, now available via API. The model targets long-horizon agentic tasks and strengthens visual understanding and agentic coding capabilities, covering complex coding projects as well as visual coding scenarios such as creative design and 3D. The company says it can be compared with GPT-5.5, GPT-5.6-Sol, Claude-Opus-4.8 and Claude-Opus-5.
Nov 2
Meituan LongCat
Meituan LongCat Releases LongCat-2.5 Multimodal Agent Model with 1.6T Total Parameters and ~48B Activated
Meituan LongCat has released LongCat-2.5, its latest native multimodal agent model, with 1.6T total parameters and approximately 48B activated parameters. The model targets long-horizon agent tasks and enhances visual understanding and agentic coding capabilities, covering complex coding projects as well as visual coding work such as creative design and 3D. For more information, see https://t.co/ULLDweplqR.
Zhidongxi
Meituan Releases LongCat-2.5: A 1.6-Trillion-Parameter Native Multimodal Model Strengthening Long-Horizon Agent Capabilities
The Meituan LongCat team has released LongCat-2.5, a native multimodal model with a total parameter count of 1.6 trillion and approximately 48 billion parameters activated per Token. Building on LongCat-2.0, it strengthens native visual understanding and agentic coding capabilities. In benchmarks published by Meituan such as AutomationBench and LongCatClawBench, LongCat-2.5 leads Claude-Opus-5 and GPT-5.6-Sol, but on Terminal Bench v4.0 it scores 38.5, surpassing GPT-5.6-Sol while still trailing Claude-Opus-5. In response to domestic accelerators having less than 80GB of memory per card, the team reorganized parallel strategies around supernodes, claiming that single-card throughput on long-sequence SFT tasks improved by 86% over the DDP baseline.
Phoenix TechnologySelected
Meituan LongCat Team Releases 1.6-Trillion-Parameter Natively Multimodal Model LongCat-2.5
The Meituan LongCat team has released LongCat-2.5, a natively multimodal model designed for long-horizon agent tasks, with a total parameter count of 1.6 trillion and approximately 48 billion parameters activated per Token. On benchmarks published by Meituan such as AutomationBench and LongCatClawBench, LongCat-2.5 leads Claude-Opus-5 and GPT-5.6-Sol; on Terminal Bench v4.0 it scores 38.5, surpassing GPT-5.6-Sol but trailing Claude-Opus-5. The model converts visual signals into multi-level discrete Tokens and processes them uniformly with text, and it redesigns its parallelism strategy for domestic accelerators whose single-card memory is under 80GB, raising single-card throughput for long-sequence SFT by 86% over the DDP baseline.
Leiphone
Meituan Releases LongCat-2.5 Large Model, 1.6 Trillion Parameters Trained Entirely on Domestic Compute
On November 2, Meituan released its new-generation large model LongCat-2.5, a native multimodal model designed for long-horizon agent tasks with 1.6 trillion total parameters. The model continues the research direction of the LongCat series and was trained entirely on domestic compute, making it the largest training task completed on domestic compute to date.
China.com TechnologySelected
Meituan Releases LongCat-2.5 Large Model with 1.6 Trillion Total Parameters
Meituan has released its new-generation large model LongCat-2.5, a natively multimodal model designed for long-horizon agent tasks, with 1.6 trillion total parameters, trained entirely on domestic compute. Evaluation results show that the model's scores on multiple mainstream benchmarks fall between Claude-Opus-4.8 and Claude-Opus-5, and it surpasses Claude-Opus-5 on some evaluations. Meituan has also launched an AI business operations workbench called "Miaoshou" based on the LongCat series, helping local-life merchants understand industry metrics and brand rules, and call tools within authorized scope to complete business tasks.
Sina Tech
Meituan Releases LongCat-2.5 Large Model, 1.6 Trillion Parameters Trained Entirely on Domestic Compute
Meituan has released its next-generation large model LongCat-2.5, a natively multimodal model designed for long-horizon agent tasks, with 1.6 trillion total parameters, trained entirely on domestic compute. During the mid-training stage, the model conducted long-context training with sequences of up to 1M tokens, with cumulative training volume reaching trillions of tokens, and it optimized task division around supernodes to address the challenge of small per-card memory in domestic compute. Meituan says its results on multiple mainstream benchmarks fall between Claude-Opus-4.8 and Claude-Opus-5, and it surpasses Claude-Opus-5 on some evaluations.
English text is being prepared.
Meituan LongCat releases LongCat-2.5 native multimodal model, 1.6T parameters targeting long-horizon agent tasks
Meituan LongCat releases the native multimodal model LongCat-2.5, with a total parameter count of 1.6 trillion and approximately 48 billion parameters activated per token, strengthening native visual understanding and agentic coding capabilities on the basis of LongCat-2.0. The model incorporates more long-horizon trajectory data into long-sequence training, and through large-scale long-horizon reinforcement learning enhances end-to-end engineering capabilities such as complex codebase development and complete project construction, while representing visual signals as multi-level discrete tokens to uniformly process visual and textual inputs. The official statement says it has optimized long-context training and W4A4 inference on domestic compute, and it will soon be available on GitHub and HuggingFace.
Phoenix Technology
Meituan LongCat releases the LongCat-2.5 native multimodal model, with 1.6T parameters targeting long-horizon agent tasks
Meituan LongCat has released its native multimodal model LongCat-2.5, with a total parameter count of 1.6 trillion and approximately 48 billion parameters activated per token, strengthening native visual understanding and agentic coding capabilities on the basis of LongCat-2.0. The model incorporates more long-horizon trajectory data into long-sequence training and, through large-scale long-horizon reinforcement learning, improves end-to-end engineering capabilities such as complex codebase development and complete project construction. LongCat-2.5 represents visual signals as multi-level discrete tokens, enabling the backbone model to process visual and textual inputs in a unified token form and supporting tasks such as creative design and 3D.
Heat trend for this story
- Current
- 20
- Period peak
- 65 Nov 3 08:00
- 24-hour change
- -63%
Move the pointer or select the chart to inspect hourly heat. Use the left and right arrow keys with a keyboard.