Best AI models for coding.

Which models write working code: independent coding evaluations combined with blind votes on web apps the models built themselves. As of 2 Oct 2026, Claude Fable 5.1 leads for coding, ahead of GPT-6 Astra and Claude Opus 5, out of 42 ranked models. The highest-ranked open-weights model is Kimi K3 (5th). The cheapest model in the top ten is Muse Spark 1.3, at A$2.88 per million tokens.

42 MODELS RANKEDA$1 = US$0.6949UPDATED 2 OCT 2026
CODING RANKING42 MODELS · 2 OCT 2026
42 MODELS
Best AI models for coding, ranked, as of 2 Oct 2026
MODELIN AUSTRALIA
01Claude Fable 5.1ANTHROPIC95.281.6175191.4%A$28.78GLOBAL
02GPT-6 AstraOPENAI90.576.9178988.4%A$28.78GLOBAL
03Claude Opus 5ANTHROPIC85.478.0169489.1%A$14.39IN AU
04Qwen3.8 MaxALIBABA80.976.2167188.8%A$4.32NOT IN AU
05Kimi K3MOONSHOT AI80.076.2165885.0%A$8.63GLOBAL
06Muse Spark 1.3META79.275.8165584.3%A$2.88NOT IN AU
07GPT-5.6 SolOPENAI79.177.4161988.0%A$11.51GLOBAL
08Grok 4.6XAI78.276.8162088.4%A$4.32GLOBAL
09Claude Fable 5ANTHROPIC78.176.5162684.6%A$28.78GLOBAL
10GLM-5.3Z.AI75.274.8162283.9%A$3.09SELF-HOST
11Gemini 3.7 FlashGOOGLE75.176.1159285.8%A$2.16GLOBAL
12Gemini 3.8 FlashGOOGLE74.776.3158387.6%A$2.16GLOBAL
13Qwen3.8 Flash NextALIBABA73.773.1163886.1%A$0.33SELF-HOST
14GPT-5.6 TerraOPENAI70.876.7152088.0%A$6.48GLOBAL
15Claude Opus 4.8ANTHROPIC69.774.3155684.6%A$14.39IN AU
16GLM-5.3 FlashZ.AI69.571.5161584.3%A$0.34SELF-HOST
17Claude Opus 4.7ANTHROPIC68.673.6155783.1%A$14.39IN AU
18GPT-5.5OPENAI67.474.9151284.3%A$16.19GLOBAL
19Grok 4.5XAI66.472.4155281.6%A$4.32NOT IN AU
20Muse Spark 1.2META64.672.2153280.1%A$2.88NOT IN AU
21GLM-5.2Z.AI64.568.8160577.9%A$3.09SELF-HOST
22Claude Sonnet 5ANTHROPIC64.171.5154080.5%A$5.76IN AU
23Muse Spark 1.1META63.971.3154277.9%A$2.88
24DeepSeek V4 FlashDEEPSEEK63.369.1158178.7%A$0.95GLOBAL
25DeepSeek V4 ProDEEPSEEK62.968.8158278.7%A$2.85GLOBAL
26GPT-5.6 LunaOPENAI62.471.4151880.9%A$0.65GLOBAL
27Qwen3.8 27BALIBABA62.468.1159079.8%A$1.62SELF-HOST
28Gemini 3.6 FlashGOOGLE60.269.2153677.5%A$2.16GLOBAL
29Gemini 3.5 FlashGOOGLE59.070.1149978.7%A$4.86IN AU
30GPT-5.4OPENAI58.171.1146578.3%A$8.09IN AU
31Qwen3.7 MaxALIBABA53.766.0151574.5%A$5.40
32Gemini 3.1 ProGOOGLE53.168.8144673.8%A$6.48GLOBAL
33Claude Sonnet 4.6ANTHROPIC49.463.0152271.2%A$8.63IN AU
34Kimi K2.6MOONSHOT AI46.661.8150965.9%A$2.46SELF-HOST
35Hy3TENCENT41.958.8150864.4%A$0.35SELF-HOST
36MiMo V2.5 ProXIAOMI41.860.2147765.2%A$0.78SELF-HOST
37Muse SparkMETAPARTIAL39.858.662.2%
38MiniMax M3MINIMAX39.758.6148265.2%A$0.76SELF-HOST
39GPT-5.4 miniOPENAI29.656.1139759.2%A$2.43
40Gemini 3.5 Flash-LiteGOOGLE22.049.3144053.6%A$1.22GLOBAL
41Gemma 4 31BGOOGLE7.343.4136543.4%A$0.29SELF-HOST
42Mistral Medium 3.5MISTRAL AI5.546.9126350.6%A$4.32SELF-HOST
HOW THIS RANKING WORKS

What goes into the coding score.

The Coding Index covers agentic terminal work and scientific programming; the WebDev arena adds how people rate finished, running apps. Terminal-Bench is shown alongside because it is the closest test to how coding agents are actually used, but it is already inside the Coding Index, so it isn’t counted twice.

60%AA Coding Index

Artificial Analysis’s composite of its coding evaluations, including Terminal-Bench and SciCode.

40%LMArena WebDev

Blind votes on working web apps each model builds from the same prompt.

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 coding.

As of 2 Oct 2026, Claude Fable 5.1 ranks first, followed by GPT-6 Astra and Claude Opus 5. The ranking combines AA Coding Index and LMArena WebDev and is refreshed weekly.

Kimi K3 from Moonshot AI is the highest-ranked open-weights model (5th of 42). Open weights can be run on your own infrastructure, including in an Australian region.

Of the top ten, Muse Spark 1.3 is cheapest at A$2.88 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.

The Coding Index covers agentic terminal work and scientific programming; the WebDev arena adds how people rate finished, running apps. Terminal-Bench is shown alongside because it is the closest test to how coding agents are actually used, but it is already inside the Coding Index, so it isn’t counted twice. Each input is rescaled 0–100 across the models we track, then weighted: AA Coding Index 60% and LMArena WebDev 40%.

PUT THE COMPARISON TO WORK

Need help choosing and using AI for your business?