VerdictPal
Join

Models / Meituan

LongCat-2.0

LongCat · Released 2026-06-29

Meituan's June 2026 agentic-coding model: a 1.6T-parameter MoE with roughly 48B active parameters per token, 1M-context training, MIT licensing, and a public reveal as the model behind OpenRouter's Owl Alpha stealth route.

Track as a near-frontier open-weights coding contender, but wait for weights and independent reruns before treating the OpenRouter/Owl Alpha adoption story as durable evidence.

#33 of 51 on AA Index · current snapshot
34 AA Index · 10 benchmark rows · 2 sourcesPosition is benchmark-specific — not a cross-family or cross-source ranking.
Composite47
Context1M
Input / 1M$0.75
Output / 1M$3.0
Knowledge cutoffopen
Output speed0 t/s
TTFT (default API)0.00s
Statussolid
OpenRouter
ReasoningCodingAgenticLong context
Reasoning
78
Coding
60
Agentic
71
Long context
58

67average across 4 tested families

Editor's note
LongCat-2.0 is important because the adoption happened before the brand reveal: Owl Alpha had already become a high-volume OpenRouter coding route. Keep the card sober: Meituan's 1.6T / 48B-active / 1M-context claims are official, but the weights are still 'coming soon' and benchmark claims remain vendor-reported.
Benchmark placements

Where LongCat-2.0 places on each public benchmark source that publishes a row. Rank counts every model with a latest row in the same test — not a universal quality score.

BenchmarkFamilyRankScoreSourceDate
Terminal-BenchAgentic#21/ 6971Meituan LongCat-2.0 release materials2026-06-29
SWE-bench VerifiedCoding#23/ 2360Meituan LongCat-2.0 release materials2026-06-29
Humanity's Last ExamReasoning#33/ 6532Artificial Analysis2026-07-21
Artificial Analysis Intelligence IndexReasoning#47/ 6734Artificial Analysis2026-07-21
AA-LCRLong context#50/ 6458Artificial Analysis2026-07-21
τ³-Bench BankingAgentic#52/ 6413Artificial Analysis2026-07-21
GPQA DiamondReasoning#53/ 6978Artificial Analysis2026-07-21
SciCodeCoding#57/ 6535Artificial Analysis2026-07-21
AA output speedPerformance#66/ 69Artificial Analysis2026-07-21
AA time to first tokenPerformance#66/ 69Artificial Analysis2026-07-21
Family context

The three highest-scoring carded models in each capability family. Where LongCat-2.0 shows up, it's highlighted.

ReasoningLongCat-2.0 · 78
  1. 1
    Claude Mythos Preview
    95
  2. 2
    GPT-5.6 Sol
    94
  3. 3
    Kimi K3
    94
CodingLongCat-2.0 · 60
  1. 1
    GPT 5.5
    96
  2. 2
    Claude Opus 4.8
    96
  3. 3
    Claude Opus 4.6
    96
AgenticLongCat-2.0 · 71
  1. 1
    GPT 5.5
    100
  2. 2
    Claude Opus 4.8
    100
  3. 3
    Claude Fable 5
    100
Long contextLongCat-2.0 · 58
  1. 1
    Kimi K3
    75
  2. 2
    GPT-5.1
    75
  3. 3
    GPT-5.6 Sol
    74
Newest receipts

Benchmark rows added to the public ledger for LongCat-2.0 in the last 120 days. Older rows live in the full table below.

10 recent
BenchmarkFamilyScoreSourceDays ago
AA-LCRLong context58%Artificial Analysis2
AA output speedPerformance0 t/sArtificial Analysis2
AA time to first tokenPerformance0.00sArtificial Analysis2
Artificial Analysis Intelligence IndexReasoning33.5%Artificial Analysis2
GPQA DiamondReasoning78%Artificial Analysis2
Humanity's Last ExamReasoning32.1%Artificial Analysis2
SciCodeCoding35.4%Artificial Analysis2
τ³-Bench BankingAgentic12.78%Artificial Analysis2
Terminal-BenchAgentic70.8%Meituan LongCat-2.0 release materials24
SWE-bench VerifiedCoding59.5%Meituan LongCat-2.0 release materials24
Family coverage

A high composite that hides a weak family is a trap. These bars surface the families where this model hasn't been publicly tested, and where it leads.

Reasoning78
Coding60
Agentic71
Long context58
Price vs. performance
0255075100$0.10$0.30$1$3$10$30input USD / 1M tokens · log scale →← intelligence

Hover or tab any dot for name, score, and input price.

AnthropicOpenAIMistralAlibabaDeepSeekGoogleMoonshotMiniMaxxAIMetaZhipuXiaomiMeituanCohere

Full benchmark ledger

Every catalog benchmark for LongCat-2.0. Scores link to the original source; gaps mean no public row exists yet.

10 sourced rows

10 of 32 catalog benchmarks have a sourced row for LongCat-2.0.

Reasoning

Coding

Agentic

Long context

Performance

Changelog
  1. Meituan revealed LongCat-2.0 and identified OpenRouter's Owl Alpha stealth route as LongCat-backed; GitHub/Hugging Face pages say model weights are coming soon.
Sources