whatmodel.sh / LongCat 2.0 (1.6T-A48B)
LongCat 2.0 (1.6T-A48B)
↗ share
+ contribute
LongCat 2.0 (1.6T-A48B) is an open-weights language model from Meituan, released June 2026. Its published size is 1600B parameters (48B active per token) and a 1M-token context window. Licence: mit.
$ how it compares
The LLMs ranked either side of LongCat 2.0 (1.6T-A48B), out of 828 with a score, then more from Meituan.
| # | model | score | where | ||
|---|---|---|---|---|---|
| 61 |  Kimi K2.6 | 73±1 | api $1.71/M +local | ||
| 62 |  GPT-5 Pro | 73±1 | api $41.25/M | ||
| 63 | LongCat 2.0 (1.6T-A48B) | 73±2 | api $1.30/M +local | ||
| 64 |  Gemini 3 Pro (Preview) | 73±1 | comparison only | ||
| 65 | GPT-5 | 72±1 | api $3.44/M | ||
| more from Meituan | |||||
| 114 | LongCat-Flash-Thinking (560B-A27B) | 69±1 | comparison only | ||
| 157 | LongCat-Flash-Chat (560B-A27B) | 66±1 | local | ||
$ what it's good at
agent tasks▁▁▃▆█▆▅▅▅▄▃▁ #71 of 300 * · 23% of its score · 6 tests
Beats GPT-5 Pro here, though they rank higher overall. Trails MiMo-V2.5 (310B-A15B), Muse Glimmer 30B, MiniMax-M3 here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| RWSearch lab | 78.8% ✎ | ████··█····█ beats 2/5 | 13% |
| EEGAgentBench overall (tool-based EEG analysis agents, 6 tasks) | 44.5% ✎ | ▂·▁▁··▇▇█▂·▁ beats 9/25 | 10% |
| SWE-bench Multilingual lab | 77.3% ✎ | ▁··▁▃▂▅▇█▃▃▁ beats 47/60 | — |
| Terminal-Bench 2.1 lab | 70.8% ✎ | ▁·▁·▁▁▄▃▄█▃▇ beats 35/56 | — |
| BrowseComp lab | 79.9% ✎ | ▂·▄▁▂▃▆▄█▃▃▄ beats 34/42 | — |
| SWE-bench Pro lab | 59.5% ✎ | ▅▃▃·▃·▅▅▅█▃▃ beats 14/15 | — |
codingnot ranked on this task yet · 2 tests
| test | result | vs the others in its run | counts |
|---|---|---|---|
| SWE-bench Multilingual lab | 77.3% ✎ | ▁··▁▃▂▅▇█▃▃▁ beats 47/60 | — |
| SWE-bench Pro lab | 59.5% ✎ | ▅▃▃·▃·▅▅▅█▃▃ beats 14/15 | — |
chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #131 of 393 * · 4.9% of its score · 1 test
Trails LFM2.5 8B-A1B, LFM2.5 2.6B, Nemotron-Cascade-8B here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| IFEval (aggregation unspecified) lab | 90.0% ✎ | █·██·█···█·█ beats 2/5 | 4.9% |
math▁▂▃▅▆▅▄█▄▁▁▁ #57 of 398 * · 14% of its score · 1 test
Trails INTELLECT-3, GLM-4.7, A.X K2 (688B-A33B) here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| IMO-AnswerBench lab | 81.8% ✎ | ▁▁·▁▁▁▄▅█▄▂▃ beats 26/42 | 14% |
web search▂▁▂▆▄█▃▃▁▁·▁ #38 of 80 * · 13% of its score · 2 tests
Beats Gemini 3.1 Pro (Preview), Qwen3.7 Max, Gemini 3.5 Flash here, though they rank higher overall. Trails GPT-4o mini (2024-07-18), Mistral Small 3.1 24B (2503), Llama 4 Scout (109B-A17B) here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| RWSearch lab | 78.8% ✎ | ████··█····█ beats 2/5 | 13% |
| BrowseComp lab | 79.9% ✎ | ▂·▄▁▂▃▆▄█▃▃▄ beats 34/42 | — |
other tests3 not in a task · 56% of its score
| test | result | vs the others in its run | counts |
|---|---|---|---|
| GPQA Diamond lab | 88.9% ✎ | ▁▁▁▁▂▄▅▅███▄ beats 257/310 | 28% |
| FORTE (AGI-Eval) lab | 73.2% ✎ | ▄···█·····▄█ beats 2/5 | 14% |
| WritingBench lab | 83.8% ✎ | ▁·▁▂▁··▅█▂▂▁ beats 8/20 | 14% |
$ where to run it
Through an API
api · LongCat API Platform (Meituan) · $1.30/M · $0.75 in / $2.95 out per 1M tokens · 1M context · longcat.chat ↗ · price sourceUSD per 1M tokens, LongCat API Platform pay-as-you-go list price for LongCat-2.0 (uncached input / output). A limited-time discounted price ($0.30 / $1.20) was also shown and is not used; cached input $0.015/1M excluded; CNY price table not used. Currency: USD. Meituan LongCat official pricing page, observed 2026-09-25.
Other ways to run it
- Run it as published · needs ~2051.18 GB, NVIDIA only · from docs.sglang.io
sglang serve --trust-remote-code --model-path meituan-longcat/LongCat-2.0-FP8 --tp 8 --ep 8 --max-running-requests 64 --mem-fraction-static 0.92 --chunked-prefill-size 2048 --dsa-prefill-backend fa3 --kv-cache-dtype bfloat16 --model-loader-extra-config '{"enable_multithread_load":true,"num_threads":12}' --host 0.0.0.0 --port 30000
Weights and documentation: huggingface.co/meituan-longcat ↗
$ how the score was worked out · meituan-longcat/LongCat-2.0 m… 88%, EEGAgentBench: Benchmarking L… 10%
Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works
| source | results | share of the score |
|---|---|---|
 meituan-longcat/LongCat-2.0 model card · lab-reported | 6 | 88% |
 EEGAgentBench: Benchmarking LLM Agents on Short- and Long-Horizon EEG Analysis (arXiv:2609.31632v1) | 1 | 10% |
| prior · pull toward the middle | 2% | |
| weighted average | = 72.6 |
not counted (4)
| meituan-longcat/LongCat-2.0 model card | SWE-bench Pro [correct] | 59.5% | fewer than three models have this test yet, or this source ran it for this model alone |
| meituan-longcat/LongCat-2.0 model card | BrowseComp [correct] | 79.9% | fewer than three models have this test yet, or this source ran it for this model alone |
| meituan-longcat/LongCat-2.0 model card | SWE-bench Multilingual [correct] | 77.3% | fewer than three models have this test yet, or this source ran it for this model alone |
| meituan-longcat/LongCat-2.0 model card | Terminal-Bench 2.1 [correct] | 70.8% | fewer than three models have this test yet, or this source ran it for this model alone |
$ improve this page
Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.
data updated · sources and licences · json