LongCat 2.0 — Hardware Requirements & GPU Compatibility
ChatFunctions
LongCat-2.0 is Meituan's large-scale mixture-of-experts language model, with roughly 1.8 trillion total parameters and about 48 billion active per token, built for coding and agentic workflows and integrated with coding harnesses such as Claude Code and OpenClaw. Both its training and large-scale deployment run entirely on AI ASIC accelerator superpods rather than GPUs, with pretraining spanning more than 35 trillion tokens without rollbacks or loss spikes. Its key architectural addition, LongCat Sparse Attention, combines streaming-aware, cross-layer, and hierarchical indexing to speed up long-context inference, alongside a 135-billion-parameter N-gram embedding component that adds capacity outside the MoE layers. It was trained on hundreds of billions of tokens of million-token-context data. Given its scale, it needs a multi-GPU server even heavily quantized. Context length is 262,144 tokens. It is released under the MIT license, permitting unrestricted commercial and research use, and was published in July 2026.
LongCat 2.0 runs on a GPU cluster
BF16 needs ~3.9 TB · about 7 nodes of 8× H100 SXM 80 GB
MIT
Not set up for it?
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
Run LongCat 2.0
One recommended way. Every other method, OS and quant is in the step-by-step guide.
Recommended
vllm serve meituan-longcat/LongCat-2.0 --max-model-len 16384 --tensor-parallel-size 8 --pipeline-parallel-size <nodes>Size the nodes from the table above
Download: Weights on Hugging Face
Other ways
each opens step-by-stepRun in cloud
Rent 7× 8-GPU H100 SXM nodes · BF16
- Estimated hourly price
- ~$151/hr
- est., whole cluster
- GPUs
- 56 GPUs
- 4.5 TB of VRAM
How Much VRAM Does LongCat 2.0 Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | Runs on | |
|---|---|---|---|---|
| 3906.2 GB | 3931.2 GB | 16-bit · bfloat — Brain floating point 16 — preferred for training | Runs on 7 nodes of 8× H100 SXM |
Selected BF16
Brain floating point 163917.9GBat 16K context
- Weights
- 3904.3 GB
- KV cache
- 13.3 GB
- Overhead
- 0.3 GB
Server-class: ~3.9 TB at 16K context · about 7 nodes of 8× H100 SXM 80 GB
Which Hardware Can Run LongCat 2.0?
BF16 · 3906.2 GBLongCat 2.0 (BF16) requires 3906.2 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 5079+ GB is recommended. No consumer GPU has enough memory.
Needs server-class hardware
LongCat 2.0 runs on a GPU cluster · BF16 needs ~3.9 TB
| Hardware | Quant | Memory | Context | Hardware page | |
|---|---|---|---|---|---|
| GPU cluster · 7 nodes of 8× H100 SXM 80 GB | BF16 | 3917.9 GB | 64K ctx | H100 SXM → | BF16 · 3917.9 GB · 64K ctx |
By VRAM tier · typical GPUs and Macs at BF16
Consumer GPUs (up to 32 GB) and Macs up to 512 GB: too small for this model (needs ≥ 3.9 TB)8 GB tier →12 GB tier →16 GB tier →24 GB tier →Macs →
Similar models
1| Model | Score | VRAM | tok/s | Why it is similar |
|---|---|---|---|---|
| DeepSeek V4 Pro 0813DeepSeek | 129.2 | 990.8 GB | —needs > 24 GB | same size |
Specifications
- Publisher
- Meituan
- Family
- LongCat
- Parameters
- 1775.6B
- Active parameters
- 48B per token
- Architecture
- LongcatCausalLM
- Context Length
- 262,144 tokens
- Vocabulary Size
- 163,840
- Release Date
- 2026-07-05
- License
- MIT
Frequently Asked Questions
How much VRAM does LongCat 2.0 need?
LongCat 2.0 requires 3906.2 GB of VRAM at BF16.
VRAM = Weights + KV Cache + Overhead
Weights = 1775.6B × 16 bits ÷ 8 = 3551.1 GB
KV Cache + Overhead ≈ 355.1 GB (at 2K context + ~0.3 GB framework)
Fit ratings and hardware model lists check this model with room for a 16K-token context, which needs a little more memory.
VRAM usage by quantization
Can NVIDIA GeForce RTX 5090 run LongCat 2.0?
No — LongCat 2.0 requires at least 3906.2 GB at BF16, which exceeds the NVIDIA GeForce RTX 5090's 32 GB of VRAM.
Can I run LongCat 2.0 on a Mac?
LongCat 2.0 requires at least 3906.2 GB at BF16. None of the 146 Macs we list has that much usable memory (the largest has 508.5 GB after macOS overhead).
Can I run LongCat 2.0 locally?
Yes — LongCat 2.0 can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
What's the download size of LongCat 2.0?
At BF16, the download is about 3551.12 GB.
Which GPUs can run LongCat 2.0?
No single consumer GPU has enough VRAM to run LongCat 2.0 at BF16 (3906.2 GB). Multi-GPU or professional hardware is required.
Which devices can run LongCat 2.0?
LongCat 2.0 requires at least 3906.2 GB at BF16, which exceeds the usable memory of every unified-memory device we list (the largest has 508.5 GB).