🤖Live agentic ranking · AA Intelligence Index v4.1 · Updated June 2026

Best Agentic AI Models
2026 ranking for tool use and autonomous work

Rankings here prioritize Agentic Index, terminal tasks, tool-use benchmarks, and cost per Intelligence Index task where available, so the order reflects real workflow tradeoffs instead of generic chat quality alone.

What changed in the agentic index

Harder agent tasks

The ranking now leans harder into Terminal-Bench, τ-Bench, and GDPval-AA style workflows that test longer agent trajectories.

Cost per task

Token price alone misses long-horizon agent cost. Cost per Intelligence Index task gives a better unit-economics view.

Cache-aware pricing

Cached input pricing now matters for agent loops with repeated context, especially when prompts and tool traces get large.

Agentic fit lab

Find the right agent model

Task

Autonomy needed

Context load

Optimize for

Best fit

GPT-5.6 Sol (max)

OpenAI

Fit

95

Agentic

57.8

Cost / task

$0.953

Response

2.0m

Context

1.0M

Codebase agentStrong autonomyWorking context

Value route

GLM-5.3-Flash

Z AI

Fit

95

Task cost

$0.087

Response

51.3s

Good balance across score and economics

Fast route

GLM-5.3 (max)

Z AI

Fit

95

Task cost

$0.683

Response

34.3s

Good balance across score and economics

Open route

GLM-5.3-Flash

Z AI

Fit

95

Task cost

$0.087

Response

51.3s

Good balance across score and economics

Frontier map

Agentic Index vs task cost

39 high-fit models

160 with cost/task

15233240485765$0.010$0.030$0.100$0.300$1.00$3.00Agentic IndexCost per taskGPT-5.6 Sol (max) · Agentic 57.8 · $0.953 per taskGPT-5.6 Sol (xhigh) · Agentic 53.6 · $0.628 per taskGLM-5.3-Flash · Agentic 58.2 · $0.087 per taskGLM-5.3 (max) · Agentic 59.1 · $0.683 per taskQwen3.8-Flash-Next · Agentic 56.4 · $0.096 per taskGrok 4.6 (high) · Agentic 58.7 · $0.937 per taskGPT-5.6 Sol (high) · Agentic 50.6 · $0.427 per taskGrok 4.6 (medium) · Agentic 56.3 · $0.783 per taskClaude Opus 5 (Adaptive Reasoning, High Effort) · Agentic 56.1 · $1.23 per taskClaude Opus 5 (Adaptive Reasoning, Xhigh Effort) · Agentic 58.4 · $1.80 per taskGrok 4.6 (xhigh) · Agentic 56.6 · $1.23 per taskQwen3.8 2.4T A95B · Agentic 57.1 · $0.807 per taskQwen3.8 Max · Agentic 58.4 · $0.913 per taskKimi K3 (max) · Agentic 54.3 · $0.838 per taskClaude Opus 5 (Adaptive Reasoning, Max Effort) · Agentic 59.2 · $2.34 per taskClaude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback) · Agentic 56.6 · $3.14 per taskDeepSeek V4 Flash Vision (Reasoning, Max Effort) · Agentic 52.9 · $0.116 per taskGPT-5.6 Sol (medium) · Agentic 47.9 · $0.290 per taskClaude Opus 5 (Adaptive Reasoning, Medium Effort) · Agentic 50.4 · $0.724 per taskGPT-5.6 Terra (max) · Agentic 50.2 · $0.526 per taskQwen3.8 27B (xhigh) · Agentic 50.9 · $0.369 per taskClaude Opus 4.8 (Adaptive Reasoning, Max Effort) · Agentic 49.4 · $2.03 per taskDeepSeek V4 Pro 0813 (Reasoning, Max Effort) · Agentic 49.6 · $0.265 per taskGPT-5.6 Terra (xhigh) · Agentic 46.5 · $0.318 per taskGrok 4.5 (high) · Agentic 48.9 · $0.426 per taskDeepSeek V4 Flash 0731 (Reasoning, Max Effort) · Agentic 48.4 · $0.112 per taskGPT-5.5 (xhigh) · Agentic 47.4 · $1.19 per taskMuse Spark 1.2 (xhigh) · Agentic 49.3 · $0.399 per taskGPT-5.5 (high) · Agentic 45.9 · $0.816 per taskGrok 4.6 (low) · Agentic 47.8 · $0.255 per taskGLM-5.2 (max) · Agentic 45.7 · $0.445 per taskGemini 3.7 Flash (high) · Agentic 45.1 · $0.402 per taskGemini 3.7 Flash (medium) · Agentic 45.1 · $0.263 per taskGPT-5.6 Luna (max) · Agentic 46.9 · $0.049 per taskGPT-5.6 Terra (high) · Agentic 43.5 · $0.228 per taskClaude Opus 4.7 (Adaptive Reasoning, Max Effort) · Agentic 46.3 · $2.23 per taskQwen3.8 27B (medium) · Agentic 49.8 · $0.284 per taskGPT-5.6 Sol (low) · Agentic 41.6 · $0.180 per taskGPT-5.6 Luna (xhigh) · Agentic 44.4 · $0.033 per taskClaude Sonnet 5 (Adaptive Reasoning, Max Effort) · Agentic 49.7 · $1.72 per taskGPT-5.4 (xhigh) · Agentic 44.2 · $1.12 per taskGemini 3.7 Flash (low) · Agentic 41.6 · $0.165 per taskClaude Opus 5 (Adaptive Reasoning, Low Effort) · Agentic 42.1 · $0.425 per taskGPT-5.5 (medium) · Agentic 39.2 · $0.508 per taskGemini 3.6 Flash (high) · Agentic 40.5 · $0.344 per taskGemini 3.5 Flash (high) · Agentic 39.7 · $0.693 per taskMuse Spark 1.1 (xhigh) · Agentic 39.7 · $0.292 per taskGPT-5.6 Luna (high) · Agentic 41 · $0.022 per taskQwen3.8 27B (low) · Agentic 43.7 · $0.265 per taskGPT-5.6 Terra (medium) · Agentic 39.1 · $0.123 per taskClaude Sonnet 4.6 (Adaptive Reasoning, Max Effort) · Agentic 42.1 · $1.22 per taskKimi K3 (low) · Agentic 39.6 · $0.242 per taskDeepSeek V4 Pro (Reasoning, Max Effort) · Agentic 37.8 · $0.050 per taskMiniMax-M3 · Agentic 36.1 · $0.139 per taskGPT-5.6 Sol (Non-reasoning) · Agentic 36 · $0.180 per taskDeepSeek V4 Pro (Reasoning, High Effort) · Agentic 35.3 · $0.043 per taskClaude Sonnet 5 (Non-reasoning, High Effort) · Agentic 34.2 · $0.417 per taskQwen3.7 Max · Agentic 30.9 · $0.679 per taskGPT-5.5 (low) · Agentic 31.7 · $0.262 per taskGPT-5.6 Terra (low) · Agentic 31.5 · $0.096 per taskGPT-5.4 mini (xhigh) · Agentic 31.5 · $0.495 per taskInkling (xhigh) · Agentic 34.1 · $0.339 per taskDeepSeek V4 Flash (Reasoning, Max Effort) · Agentic 33.7 · $0.054 per taskSolar Pro 4 · Agentic 33.6 · $0.303 per taskKimi K2.7 Code · Agentic 30.3 · $0.222 per taskHy3 · Agentic 31.4 · $0.036 per taskKimi K2.6 · Agentic 31.2 · $0.389 per taskMiMo-V2.5-Pro · Agentic 29.5 · $0.034 per taskInkling Small · Agentic 31.9 · $0.073 per taskGPT-5.6 Luna (medium) · Agentic 31.8 · $0.012 per taskGLM-5.1 (Reasoning) · Agentic 30.6 · $0.287 per taskGPT-5.4 nano (xhigh) · Agentic 29.7 · $0.149 per taskGemini 3.1 Pro Preview · Agentic 23 · $0.335 per taskGPT-5.6 Terra (Non-reasoning) · Agentic 30.1 · $0.102 per taskLing 3.0 Flash · Agentic 29.3 · $0.038 per taskDeepSeek V4 Flash (Reasoning, High Effort) · Agentic 30.3 · $0.053 per taskQwen3.6 Plus · Agentic 29 · $0.357 per taskGrok Build 0.1 0616 · Agentic 28.9 · $0.225 per taskGPT-5.5 (Non-reasoning) · Agentic 26.1 · $0.208 per taskQwen3.8 27B (Non-reasoning) · Agentic 30.4 · $0.426 per taskGPT-5.6 Sol (max)GPT-5.6 Sol (xhigh)GLM-5.3-Flash

GPT-5.6 Sol (max)

OpenAI

Agentic

57.8

Task cost

$0.953

Shortlist

1

GPT-5.6 Sol (max)

57.8 agentic score

95

fit

2

GPT-5.6 Sol (xhigh)

53.6 agentic score

95

fit

3

GLM-5.3-Flash

58.2 agentic score

95

fit

4

GLM-5.3 (max)

59.1 agentic score

95

fit

5

Qwen3.8-Flash-Next

56.4 agentic score

94

fit

Top 3 agentic models

🥇

Agentic Index

59.2

Claude Opus 5 (Adaptive Reasoning, Max Effort)

Anthropic

Cost / task$2.34
Terminal-BenchN/A
τ-BenchN/A
GDPval-AA1822
🥈

Agentic Index

59.1

GLM-5.3 (max)

Z AI

Cost / task$0.683
Terminal-BenchN/A
τ-BenchN/A
GDPval-AA1763
🥉

Agentic Index

58.7

Grok 4.6 (high)

SpaceXAI

Cost / task$0.937
Terminal-BenchN/A
τ-BenchN/A
GDPval-AA1729

Full agentic model ranking

The models below are ranked for autonomous execution, tool use, multi-step reliability, and task-level economics.

RankModelAgenticCost / taskTerminal-Benchτ-BenchGDPval-AAResponse
1

Claude Opus 5 (Adaptive Reasoning, Max Effort)

Anthropic

59.2$2.34N/AN/A182251.9s
2

GLM-5.3 (max)

Z AI

59.1$0.683N/AN/A176334.3s
3

Grok 4.6 (high)

SpaceXAI

58.7$0.937N/AN/A172951.6s
4

Qwen3.8 Max

Alibaba

58.4$0.913N/AN/A17242.1m
5

GLM-5.3-Flash

Z AI

58.2$0.087N/AN/A176451.3s
6

GPT-5.6 Sol (max)

OpenAI

57.8$0.95366%85%17112.0m
7

Qwen3.8 2.4T A95B

Alibaba

57.1$0.807N/AN/A17151.8m
8

Qwen3.8-Flash-Next

Alibaba

56.4$0.096N/AN/A173937.2s
9

Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback)

Anthropic

56.6$3.1463%99%17231.5m
10

Kimi K3 (max)

Kimi

54.3$0.838N/AN/A16701.3m

How to choose a model for agents

If your agents browse, call APIs, run tools, or plan several steps ahead, start here rather than on a generic leaderboard. The best agentic model is the one that stays reliable under execution at a cost and response-time profile you can actually ship.

Once you have a shortlist, open the finalists on Compare and validate whether the provider you want can deliver the right price and latency profile.

Related live rankings

Agentic performance overlaps with coding and long-context performance, but it is not the same thing. Use the related pages below if your use case is specialized around software work, model routing, or large-document reasoning.

Best LLM for CodingLargest Context Window LLMBest AI Models

Ranking library

Current live rankings

Focused rankings for the decisions engineers actually make.

01OverallBest AI ModelsQuality, price, speed, and context in one live ranking.Open ranking 02CodingBest LLM for CodingLiveCodeBench, Terminal-Bench, SciCode, and engineering fit.Open ranking 03OpenBest Open Source LLMOpen-weight models for self-hosting and lower-cost deployment.Open ranking 04LocalBest Local LLMModels mapped to Mac, RTX, workstation, and server hardware.Open ranking 05Local codingBest Local LLM for CodingCoding strength filtered through realistic hardware limits.Open ranking 06OllamaBest Ollama ModelsLow-friction local models for coding, reasoning, and chat.Open ranking 07ContextLargest Context Window LLMLong documents, codebases, and retrieval-heavy work.Open ranking

Frequently Asked Questions

What is the best AI model for agents in 2026?

The live top-ranked model on this page is the best starting point. The right answer depends on whether you prioritize raw capability, reliability under tool use, or latency in production. Check the ranking table above for the current leader.

What makes an LLM agentic?

Agentic models can maintain plans across many steps, call tools reliably, follow complex instructions, and recover gracefully when a workflow hits an unexpected state.

Are coding models automatically good for agents?

Not always. Coding strength helps with tool-writing and structured output, but agentic performance also requires strong planning, tool orchestration, and error recovery.

How should I validate an agentic model shortlist?

Use benchmarks to get a shortlist, then run the finalists on the actual tasks your agent will perform. Production testing on real workflows is the only way to make the final call.

Does latency matter for agentic AI?

Yes — significantly. In multi-step agents, slow thinking compounds across many tool calls. A model that is 30% slower can double the wall-clock time of a complex workflow.

What benchmarks predict agentic performance?

Terminal-Bench, τ-Bench, and GDPval-AA are strong predictors of real agentic capability. IFBench remains useful as a supplemental instruction-following signal, but it is no longer the main filter for frontier models.