LongCat-2.5: 1.6T Parameters, 1M Context, $0.30 API

The most interesting AI lab of 2026 might be a food-delivery company. Meituan — the Chinese platform that delivers more meals than almost anyone on Earth — has spent two years quietly building a frontier model programme, and its LongCat team made LongCat-2.5 generally available: a 1.6-trillion-parameter, natively multimodal model with a 1-million-token context window, priced at $0.30 per million input tokens. That is frontier scale at budget-model prices, aimed squarely at the longest, messiest job in AI right now: autonomous agents that work inside real codebases for hours.

This guide covers what was actually released, the architecture change that matters (vision turned into tokens), Meituan's benchmark claims read honestly — including the rows where a rival wins on Meituan's own table — the pricing with its discount catch, how to try the model today, and the two things it is not yet: open weights, and independently tested.

LongCat-2.5 reached general availability on the LongCat API Platform following a preview build. The positioning shifted between versions. LongCat-2.0, which Meituan open-sourced with MIT-licensed weights, was an agentic coding model. The 2.5 pitch is broader: a model for long-horizon agentic tasks — work that spans terminals, browsers, graphical interfaces, spreadsheets and design tools, and whose unit of success is not a good answer but a finished project.

Three changes headline the release notes. First, vision became native. Second, the reinforcement learning got longer: Meituan describes training on synthetic environments mined from real GitHub issues and pull requests, with fail-to-pass verification, asynchronous rollouts, and mixed training across different agent harnesses. Third, the infrastructure absorbed the cost: the model was trained and is served on domestic Chinese accelerators, with W4A4 inference (4-bit weights and 4-bit activations) and engineering work on keeping training and inference numerically consistent.

Core Specifications Overview

Two of those numbers deserve a pause. The 48-billion active count is how a 1.6-trillion-parameter model can be served at $0.30 a million: only a small slice of the network fires for any given token, so the compute bill resembles a mid-size model's while the knowledge capacity resembles a giant's. And the 1M context is not a stretched version of a smaller window — the LongCat line was built around sparse attention designed for million-token sequences. Developers utilizing robust infrastructure aggregators like n1n.ai can easily route requests to leverage these massive context windows without changing core architecture.

The Evolution of the LongCat Program

The logic behind Meituan's infrastructure is straightforward. Meituan operates one of the highest-volume consumer platforms on Earth — ordering, dispatch, customer service, merchant tooling — and every one of those workloads is now an AI workload. By accessing reliable API aggregators like n1n.ai, developers worldwide can tap into similar high-throughput capabilities without building out localized supply chains.

Native Multimodal Architecture

Most multimodal models bolt a vision encoder onto a language model and translate between the two. LongCat-2.5 instead converts visual signals into multi-level discrete tokens — images and video frames are quantized into the same kind of token IDs the language backbone already processes, at native resolution. One sequence type, one backbone, no separate visual pathway to fine-tune around.

Why should a developer care about an internal representation choice? Because it changes what the model can be asked to do in one pass. A screenshot of a broken UI, a photo of a whiteboard, or a short clip of a failing build becomes part of the same working memory as the code and the error log.

Benchmark Analysis

Every score in this section is Meituan-reported, as no independent body has scored 2.5 yet. However, the internal deltas show where the training went:

The pattern is consistent: the largest jumps sit exactly where Meituan said the reinforcement learning went — long, verified, multi-step engineering tasks (DeepSWE up 55 points, Terminal-Bench v4.0 up from near-zero) rather than chat quality. When integrating models like LongCat-2.5 into enterprise pipelines, developers often turn to n1n.ai for unified billing, low latency, and fallback support.

Pricing Structure and Economics

The pay-as-you-go prices, per million tokens, at the launch's limited-time discount:

Three details make this better than the headline. Cached input costs 2% of uncached — for a coding agent that resends the same repository context every turn, most of the input bill effectively drops fifty-fold. Only successful calls are billed (HTTP 200s), and prepaid balance does not expire. There is no length tiering: a near-million-token request pays the same rates as a short one.

How to Integrate LongCat-2.5 Today

1. Through the API platform: Create a key on Meituan's LongCat API Platform, or access it via reliable multi-model aggregators like n1n.ai. Call the model in either the OpenAI or the Anthropic request format.
2. Inside your agent harness: Meituan states compatibility with Claude Code, Codex, Hermes, OpenClaw, OpenCode and Kilo Code. Point the harness at the endpoint as the model provider.
3. Through a gateway: Vercel AI Gateway and other providers carry the model for teams outside China.

Conclusion

LongCat-2.5 is the price-performance story of late 2026: frontier scale, a genuine architecture idea in vision-as-tokens, agent training aimed at finishing work rather than chatting about it, and API pricing that makes a day of autonomous coding cost less than lunch. It launches with two honest asterisks — vendor-only benchmarks and weights that exist as a "coming soon" link — and one financial asterisk, a discount with no visible expiry. Test it now while testing is cheap, and leverage professional routing tools to streamline your deployment.

Get a free API key at n1n.ai