Meituan open-sources LongCat-2.5, 1.6T-parameter model trained without Nvidia GPUs
11/15/2026 — 11/15, 18:17·1 sources (all)·1 reports (all)
Key takeaways
- Meituan has released the weights of LongCat-2.5, an agent model with 1.6 trillion total parameters and about 48 billion activated per token,…
Background & analysis
According to a report published by The Decoder at 18:06 on November 15, 2026, Meituan has released the weights of LongCat-2.5, a model built for AI agents. The model has 1.6 trillion parameters in total, with roughly 48 billion activated per token. Meituan states that training was carried out entirely on Chinese hardware, and the report makes no mention of Nvidia GPUs being used at any point.
On evaluation performance, Meituan says LongCat-2.5 scored 88.4 on Terminal Bench, placing it just behind Claude Opus 5. The same report notes that the sample deployment of the model runs on 194 Huawei Ascend chips. The article, titled Meituan releases a 1.6-trillion-parameter agent model that never touched an Nvidia GPU, first appeared on The Decoder, and Meituan's team behind the model is described as coming from the Chinese food delivery company.
At this stage, the publicly available details about LongCat-2.5 stop there, with no further reports yet on subsequent developments, additional benchmark results, or wider availability of the release.
AI-generated from 1 reports · updated 2 days ago
- Heat index
- 29
- All sources
- 1
- All reports
- 1
- First seen
- 2 days ago
Reports on this story headlines open the original
Meituan has released the weights for LongCat-2.5, a 1.6-trillion-parameter mixture-of-experts model built for agents and trained entirely on Chinese hardware, The Decoder reports. Meituan says it scores 88.4 on Terminal Bench, just behind Claude Opus 5, and the sample deployment runs on 194 Huawei Ascend chips.
The DecoderAI score 74
Other stories people are talking about
- Heat index 1345RisingNvidia board not discussing Jensen Huang's return, probe into his conduct advances22 sources
- Heat index 1241SurgeCalifornia Sues OpenAI Over Agents Escaping Tests, Intruding into Hugging Face15 sources
- Heat index 901Alibaba releases MoE model Qwen4-Max with 1M token context for low-cost, long-running tasks17 sources
- Heat index 695US Commerce Department issues interim final rule restricting remote access to AI chips via overseas data centers11 sources
- Heat index 532Anthropic to begin trading on Nasdaq this week under ticker ANT9 sources
How is heat calculated?About the methodHide
Heat counts how many independent sources covered a story in the last 48 hours: one source counts once no matter how many posts it published, decaying with a 24-hour half-life. What ranks first is what many people are talking about.
This page aggregates public feeds. Headlines and summaries are machine-organized and remain the property of the original authors; verify important facts at the source. If you believe a headline or summary infringes your rights, email the contact address in the footer with the page URL and basis of your claim, and we will remove or replace it after review.
- Surge
- Discussion rising fast
- New
- First report within 6 hours
- Rising
- Still gathering discussion

