LongCat 2.0 benchmarks & scores

Q: How much does LongCat 2.0 cost for agentic engineering?

The current benchmark uses a 3:1 input-to-output blend of $0.53 per million blended tokens. That places it #52 of 125 by price, where a lower rank number means less expensive.

Meituan Proprietary Released Jul 2026

—

Vibe Coding Index

Independent LongCat 2.0 AI coding benchmark results Intelligence —, Coding 45, Agentic —, with a Vibe Coding Index of —. Compare its overall rank, token price, context window, and practical agentic workflow strengths below.

Blended price #52/125

$0.53

$/M tokens · 3 to 1 in, out

Context

Window size

Evidence confidence

Pending evidence

33% core coverage

The model remains outside verified rankings until all core scores are available.

1/3

Core scores

Evidence sources

Aug 18

Last measured

Artificial analysis →

Catalog capabilities

What the model supports

1M context

Frequency penalty Include reasoning Logit bias Max tokens Min p Presence penalty Reasoning Repetition penalty Seed Stop

Input

Text

Output

Text

Knowledge cutoff

Not listed

Cache read price

$0.01 per million tokens

Suite radar

Radar hidden — some suite scores are unavailable.

Suite scores

Intelligence

General reasoning & knowledge (Intelligence Index)

—

Coding #64/139

Code generation & software tasks (Coding Index)

45.3

Agentic

Multi-step tool use & agentic workflows (Agentic Index)

—

Source record

Source. Artificial Analysis (artificialanalysis.ai) via OpenRouter (openrouter.ai/rankings).

Artificial Analysis via OpenRouter

TheVibeFather rankings

LongCat 2.0 practical agentic coding scorecard

This is our algorithmic assessment of where LongCat 2.0 ranks for common vibe coding and agentic engineering workflows. Every stable 0–10 score is calculated consistently from the sourced Intelligence, Coding, and Agentic signals above—never from an unsourced opinion score.

Overall VibeFather rating

— /10

Practical area

Our score

Field rank

Small, well-defined code changes

Partial inputs

Precision on scoped edits, fixes, and implementation tasks.

Unverified —

Field rank Unranked

UI and CSS iteration

Partial inputs

Front-end implementation with iterative tool-driven refinement.

Unverified —

Field rank Unranked

Routine debugging

Partial inputs

Diagnosing failures and turning reasoning into correct code changes.

Unverified —

Field rank Unranked

Repository-wide refactors

Partial inputs

Coordinating larger edits across files while preserving intent.

Unverified —

Field rank Unranked

Architecture decisions

Partial inputs

Reasoning through tradeoffs, constraints, and system-level choices.

Unverified —

Field rank Unranked

Tool use and workflow execution

Partial inputs

Planning and completing multi-step work with tools and feedback loops.

Unverified —

Field rank Unranked

Long autonomous coding tasks

Partial inputs

Sustaining coherent progress across longer agentic engineering runs.

Unverified —

Field rank Unranked

Overall VibeFather rating

Partial inputs

Our complete Vibe Coding Index, expressed on the same 0–10 practical scale.

Unverified —

Field rank Unranked

1 · Sourced signals

We start with Intelligence, Coding, Agentic benchmark evidence, and preserve missing values as unverified.

2 · Practical lenses

Our code-controlled algorithm blends the native 0–100 signals by workflow and expresses the result on a stable 0–10 scale. Every required input must be present.

3 · Field ranking

Each workflow score is ranked against every model with comparable evidence, so positions update when the benchmark field changes.

The Vibe Coding Index is TheVibeFather’s code-controlled 0–100 composite for practical coding and agentic capability. Price and adoption are reported independently and never alter quality. Exact weighting remains proprietary, inputs, missing-data behavior, and per-category ranks are disclosed here. Read the benchmark methodology →

Recorded score history

How LongCat 2.0 has moved

All benchmark history →

The first historical point will appear after the next completed benchmark refresh.

Where LongCat 2.0 sits

Its bar stays full-color on each index while the rest of the field recedes. The price scatter below keeps the same highlight.

Intelligence

Artificial Analysis Intelligence Index · Higher is better

63.1

Claude Opus 5 62.1

Claude Fable 5 60.9

Qwen3.8 2.4T A95B 57.3

Claude Opus 4.8 57

Muse Spark 1.2 56.8

Muse Spark 1.2 56.6

GPT-5.6 Terra 56.3

GPT-5.5 56

Gemini 3.7 Flash 55.8

Grok 4.5 55.3

Claude Sonnet 5 55

Claude Opus 4.7 53.2

DeepSeek V4 Pro 0813 53.2

Muse Spark 1.1 53.1

GPT-5.4 52.6

GLM-5.2 52.3

GPT-5.6 Luna 52

Gemini 3.5 Flash 51.8

DeepSeek V4 Flash 0731 51.6

Gemini 3.6 Flash 48.4

Claude Sonnet 4.6 47.7

Gemini 3.1 Pro Preview 46.7

Qwen3.7 Max 45.4

MiniMax-M3

Reasoning models are marked with a lightbulb Full board →

Model benchmark FAQ

LongCat 2.0 for vibe coding and agentic engineering

Answers update from this model’s current sourced scores, practical rankings, and verified pricing.

What is LongCat 2.0's Vibe Coding Index?

LongCat 2.0 has a Vibe Coding Index of unverified and is not currently ranked in the current field. The index is TheVibeFather's stable composite for practical AI coding, combining independently sourced Intelligence, Coding, and Agentic benchmark signals. Price and adoption are compared separately and never inflate the quality score.

How does LongCat 2.0 rank for vibe coding?

LongCat 2.0 does not yet have enough verified benchmark inputs for a practical vibe coding category rank.

Is LongCat 2.0 good for agentic coding and autonomous tasks?

The agentic coding score for LongCat 2.0 is still partial because one or more required benchmark signals are unverified.

What coding benchmark scores does LongCat 2.0 have?

The current sourced benchmark profile for LongCat 2.0 is Intelligence unverified, Coding 45/100, Agentic unverified. TheVibeFather converts those 0–100 signals into consistent practical 0–10 workflow scores so models can be compared for real agentic engineering work.

How much does LongCat 2.0 cost for agentic engineering?

The current benchmark uses a 3 to 1 input-to-output blend of $0.53 per million blended tokens. That places it #52 of 125 by price, where a lower rank number means less expensive.

Are vibe coding, agentic coding, and agentic engineering the same?

They describe overlapping AI-assisted software workflows. Vibe coding emphasizes directing software through intent and iteration, agentic coding emphasizes a model planning and executing multi-step work with tools, agentic engineering is the broader discipline of designing, supervising, and validating those workflows. TheVibeFather benchmarks the shared capabilities behind all three terms on every model page, including LongCat 2.0.

Compare LongCat 2.0 → Back to leaderboard