Best AI models for agents and automation.

Models ranked on multi-step work: following policy while using tools, getting jobs done in a terminal, and how people rate real agent sessions. As of 2 Oct 2026, Claude Fable 5.1 leads for agents, ahead of GPT-6 Astra and Muse Spark 1.3, out of 41 ranked models. The highest-ranked open-weights model is GLM-5.3 (8th). The cheapest model in the top ten is Gemini 3.8 Flash, at A$2.16 per million tokens.

41 MODELS RANKEDA$1 = US$0.6949UPDATED 2 OCT 2026
AGENTS RANKING41 MODELS · 2 OCT 2026
41 MODELS
Best AI models for agents and automation, ranked, as of 2 Oct 2026
MODELIN AUSTRALIA
01Claude Fable 5.1ANTHROPIC96.647.2%91.4%146A$28.78GLOBAL
02GPT-6 AstraOPENAI86.641.4%88.4%122A$28.78GLOBAL
03Muse Spark 1.3META85.050.5%84.3%42A$2.88NOT IN AU
04Grok 4.6XAI84.750.7%88.4%12A$4.32GLOBAL
05Claude Opus 5ANTHROPIC84.342.1%89.1%88A$14.39IN AU
06GPT-5.6 SolOPENAI84.144.3%88.0%71A$11.51GLOBAL
07Qwen3.8 MaxALIBABA83.747.8%88.8%28A$4.32NOT IN AU
08GLM-5.3Z.AI82.950.3%83.9%26A$3.09SELF-HOST
09Kimi K3MOONSHOT AI80.846.0%85.0%40A$8.63GLOBAL
10Gemini 3.8 FlashGOOGLE80.344.9%87.6%30A$2.16GLOBAL
11GLM-5.3 FlashZ.AI77.347.2%84.3%-3A$0.34SELF-HOST
12Claude Fable 5ANTHROPIC77.238.1%84.6%84A$28.78GLOBAL
13Qwen3.8 Flash NextALIBABA76.145.4%86.1%-8A$0.33SELF-HOST
14Qwen3.8 27BALIBABA73.948.0%79.8%-16A$1.62SELF-HOST
15GPT-5.6 TerraOPENAI73.840.2%88.0%8A$6.48GLOBAL
16GPT-5.5OPENAI73.639.0%84.3%41A$16.19GLOBAL
17Grok 4.5XAI72.042.1%81.6%11A$4.32NOT IN AU
18Claude Opus 4.8ANTHROPIC71.934.2%84.6%69A$14.39IN AU
19Claude Sonnet 5ANTHROPIC70.337.3%80.5%48A$5.76IN AU
20Claude Opus 4.7ANTHROPICPARTIAL70.034.6%83.1%A$14.39IN AU
21DeepSeek V4 FlashDEEPSEEK68.239.4%78.7%18A$0.95GLOBAL
22DeepSeek V4 ProDEEPSEEK67.939.6%78.7%13A$2.85GLOBAL
23GPT-5.4OPENAI67.239.6%78.3%9A$8.09IN AU
24GLM-5.2Z.AI64.834.6%77.9%36A$3.09SELF-HOST
25Gemini 3.5 FlashGOOGLEPARTIAL62.832.2%78.7%A$4.86IN AU
26Gemini 3.7 FlashGOOGLE62.732.8%85.8%-16A$2.16GLOBAL
27Muse Spark 1.2META59.834.8%80.1%-30A$2.88NOT IN AU
28GPT-5.6 LunaOPENAI58.531.1%80.9%-11A$0.65GLOBAL
29Claude Sonnet 4.6ANTHROPIC55.334.4%71.2%-15A$8.63IN AU
30Muse Spark 1.1META53.831.8%77.9%-46A$2.88
31Gemini 3.6 FlashGOOGLE48.629.9%77.5%-77A$2.16GLOBAL
32Kimi K2.6MOONSHOT AIPARTIAL38.923.3%65.9%A$2.46SELF-HOST
33Gemini 3.1 ProGOOGLE38.321.4%73.8%-73A$6.48GLOBAL
34GPT-5.4 miniOPENAIPARTIAL36.125.6%59.2%A$2.43
35Hy3TENCENT35.622.9%64.4%-57A$0.35SELF-HOST
36Qwen3.7 MaxALIBABA31.511.8%74.5%-52A$5.40
37MiniMax M3MINIMAX27.615.3%65.2%-67A$0.76SELF-HOST
38MiMo V2.5 ProXIAOMI21.89.9%65.2%-72A$0.78SELF-HOST
39Gemini 3.5 Flash-LiteGOOGLE13.817.5%53.6%-153A$1.22GLOBAL
40Mistral Medium 3.5MISTRAL AI13.215.1%50.6%-117A$4.32SELF-HOST
41Gemma 4 31BGOOGLEPARTIAL6.914.8%43.4%A$0.29SELF-HOST
HOW THIS RANKING WORKS

What goes into the agents score.

τ²-Bench is the closest public test to a business agent — a customer conversation where the model has tools and rules and must resolve the request correctly. Terminal-Bench covers operational tasks; LMArena’s agent arena adds human judgement of real sessions. A model needs at least two of the three to be ranked.

40%τ²-Bench (banking)

Customer-service agent conversations in a banking setting, where the model must use tools and follow policy to resolve the request.

30%Terminal-Bench 2.1

Practical tasks completed by an agent working in a real command-line environment.

30%LMArena Agent

Multi-step agent sessions rated by the people running them. LMArena publishes a relative score, so the rank is the useful part.

Scores are rescaled 0–100 across the models we track, so they compare models with each other rather than against a fixed bar. Full methodology →

QUESTIONS

Choosing a model for agents.

As of 2 Oct 2026, Claude Fable 5.1 ranks first, followed by GPT-6 Astra and Muse Spark 1.3. The ranking combines τ²-Bench (banking), Terminal-Bench 2.1 and LMArena Agent and is refreshed weekly.

GLM-5.3 from Z.ai is the highest-ranked open-weights model (8th of 41). Open weights can be run on your own infrastructure, including in an Australian region.

Of the top ten, Gemini 3.8 Flash is cheapest at A$2.16 per million tokens (blended), against A$28.78 for Claude Fable 5.1. Prices are list prices converted at A$1 = US$0.6949, excluding GST.

Of the top ten, Claude Opus 5 can run with requests kept in Australia on at least one major cloud platform. See the Australia page for each platform.

τ²-Bench is the closest public test to a business agent — a customer conversation where the model has tools and rules and must resolve the request correctly. Terminal-Bench covers operational tasks; LMArena’s agent arena adds human judgement of real sessions. A model needs at least two of the three to be ranked. Each input is rescaled 0–100 across the models we track, then weighted: τ²-Bench (banking) 40%, Terminal-Bench 2.1 30% and LMArena Agent 30%.

PUT THE COMPARISON TO WORK

Need help choosing and using AI for your business?