What is the smartest AI model in October 2026?

On the Artificial Analysis Intelligence Index, Claude Opus 5.5 scores highest of the models we track at 58, followed by Claude Sonnet 5.5 (56) and Claude Fable 5.1 (53). Claude Opus 5.5 costs $0.11 per average coding-agent request on our measured traffic.

Source: Artificial Analysis (artificialanalysis.ai). Fetched October 13, 2026. Cost per request: Standard Compute gateway data, 79,301 coding-agent requests, Sep 13 to Oct 12, 2026. How we compute this

Run current Claude, GPT, Gemini and open models in your agent from $19/month

See plans

36 current models ranked by Artificial Analysis Intelligence Index

Model	Intelligence IndexArtificial Analysis	Terminal-Bench 4.0Artificial Analysis	Cost per requestour data
1. Claude Opus 5.5max effort	58	59.6%	$0.11
2. Claude Sonnet 5.5max effort	56	63.6%	$0.053
3. Claude Fable 5.1max effort	53	52.0%	$0.24
4. GPT-6 Astramax effort	53	59.1%	$0.41
5. Gemini 4 Argonhigh effort	53	57.1%	$0.11
6. Muse Wondermax effort	52	52.0%	$0.11
7. GPT-6.1 Solmax effort	52	56.1%	$0.067
8. Grok 4.7xhigh effort	46	25.8%	$0.13
9. GLM 5.3max effort	45	41.9%	$0.057
10. Step 5 Preview	44	33.3%	$0.025
11. Claude Haiku 5.5max effort	43	32.8%	$0.013
12. GPT-5.6 Terramax effort	42	35.4%	$0.084
13. Gemini 3.8 Flashhigh effort	41	19.7%	$0.048
14. DeepSeek V4.1 Flashmax effort	39	26.8%	$0.0067
15. GPT-5.4xhigh effort	39*	n/a	$0.10
16. Mistral Large 4Mistral Large 4 Preview	38	26.8%	$0.042
17. GPT-6 Lunamax effort	38	12.6%	$0.0041
18. DeepSeek V4 ProDeepSeek V4 Pro 0813, max effort	36	14.1%	$0.030
19. GPT-5.3-Codexxhigh effort	33*	n/a	$0.055
20. Gemini 3.1 ProGemini 3.1 Pro Preview	30	4.0%	$0.095
21. Qwen3.7 Max	29	1.5%	$0.063
22. MiniMax M3	29	2.0%	$0.013
23. Kimi K2.6reasoning	27	0.5%	$0.024
24. Kimi K2.7 Code	26	1.0%	$0.027
25. Grok 4.20Grok 4.20 0309 v2, reasoning	26*	n/a	$0.068
26. Qwen3.7 Plus	25	1.0%	$0.014
27. GPT-5.4 Minixhigh effort	24	2.0%	$0.023
28. GLM 4.7reasoning	22*	n/a	$0.017
29. DeepSeek V3.2reasoning	21*	n/a	$0.013
30. MiniMax M2.1	21*	n/a	$0.0065
31. GLM 4.6reasoning	19*	n/a	$0.015
32. Gemini 3.1 Flash Lite	16	0.5%	$0.0082
33. DeepSeek R1 (0528)may '25	13*	n/a	$0.031
34. Qwen3 235B A22BQwen3 235B A22B 2507 Instruct	12*	n/a	$0.0096
35. Llama 4 Maverick	10*	0.0%	$0.0095
36. Devstral 2	9	0.0%	$0.010

Source: Artificial Analysis (artificialanalysis.ai). Each model is measured as the variant named under it. Cost per request is ours: every one of 79,301 real coding-agent requests priced at list rates (OpenRouter's listed price for some older and open models). Methodology

Older versions

Model	Intelligence IndexArtificial Analysis	Terminal-Bench 4.0Artificial Analysis	Cost per requestour data
Claude Fable 5superseded by Claude Fable 5.1	50	42.4%	$0.32
GPT-6 Solsuperseded by GPT-6.1 Sol	48	43.9%	$0.083
GPT-5.6 Solsuperseded by GPT-6.1 Sol	47	39.9%	$0.17
Claude Opus 4.8superseded by Claude Opus 5.5	42	21.7%	$0.16
GPT-5.5superseded by GPT-6.1 Sol	38	14.6%	$0.21
GPT-5.6 Lunasuperseded by GPT-6 Luna	37	11.6%	$0.0084
GLM 5.2superseded by GLM 5.3	34	1.0%	$0.057
Gemini 3.5 Flashsuperseded by Gemini 3.8 Flash	33	6.6%	$0.049
Claude Sonnet 4.6superseded by Claude Sonnet 5.5	30	3.0%	$0.097
Grok 4.3superseded by Grok 4.7	25	0.0%	$0.068
Claude Haiku 4.5superseded by Claude Haiku 5.5	17	0.0%	$0.022
Mistral Large 3superseded by Mistral Large 4	9	0.0%	$0.012

Source: Artificial Analysis (artificialanalysis.ai). Cost per request is ours. Methodology

FAQ

What does the Artificial Analysis Intelligence Index measure?

Version 4.3.2 combines 10 evaluations that Artificial Analysis runs itself: AA-Briefcase, GDPval-AA, AutomationBench-AA, Terminal-Bench 4.0, SciCode, Humanity's Last Exam, GDP.pdf, CritPt, AA-Omniscience and AA-LCR. It is a general index, so a model's rank on it can differ from its rank on coding alone.

Is the smartest model the best choice for a coding agent?

Not always. Coding agents resend their whole context on every step, so cost per request decides the bill. Claude Sonnet 5.5 is within 10 points of the top score and costs $0.053 per request on our traffic, against $0.11 for Claude Opus 5.5. Many teams run a strong model for planning and hard steps and a cheaper one for routine edits.

Where do the cost figures come from?

They are ours: 79,301 real coding-agent requests (Sep 13 to Oct 12, 2026) priced request by request at each vendor's list rates, including cache reads and long-prompt surcharges.

Current Claude, GPT, Gemini and open models are included with Standard Compute

Standard Compute

Current frontier models Claude Opus 5.5 · Claude Sonnet 5.5