Skip to content
IFAI

Model rankings

1501 models · 5 boards · 2 sources

WebDev arena · People compare web pages built by two models · Published September 11, 2026

Best web builder
gpt-6-astra-maxOpenAI1,800Score
Best open model
kimi-k3-maxMoonshot AI1,674Score
Best from China
qwen3.8-max-0902Alibaba1,681Score
Most voted
claude-haiku-4-5-20251001Anthropic27,785Votes
Top three by Score
1gpt-6-astra-maxOpenAI1,800
2claude-fable-5.1-maxAnthropic1,758
3claude-opus-5-maxAnthropic1,687
Top three open models
1kimi-k3-maxMoonshot AI · #5 overall1,674
2qwen3.8-flash-nextAlibaba · #8 overall1,635
3hy4-previewTencent · #10 overall1,624
This week's changes
Agnes 3.0 FlashIntelligence · #27New
Ling-3.0-flash-VLIntelligence · #77New
RankChangeModelScore95% intervalVotesPrice $/M
1—gpt-6-astra-maxClosedOpenAI1,800±162,281—
2—claude-fable-5.1-maxClosedAnthropic1,758±143,036—
3—claude-opus-5-maxClosedAnthropic1,687±712,087—
4—claude-opus-5-highClosedAnthropic1,660±712,566—
5—muse-spark-1.3-maxClosedMeta1,652±122,972—
6—claude-fable-5ClosedAnthropic1,628±710,081—
7—muse-spark-1.3 (xHigh)ClosedMeta1,623±142,123$2.0
8—grok-4.6-highClosedxAI1,618±104,282—
9—gpt-5.6-sol-xhigh (codex-harness)ClosedOpenAI1,617±711,916—
10—gemini-3.7-flash-highClosedGoogle1,587±123,007—
11—gemini-3.8-flash-highClosedGoogle1,568±122,732—
12—claude-opus-4-8-highClosedAnthropic1,559±713,787—
13—claude-opus-4-7ClosedAnthropic1,557±615,914—
14—claude-opus-4-7-highClosedAnthropic1,555±616,387—
15—grok-4.5ClosedxAI1,555±88,133—
16—claude-opus-4-6-highClosedAnthropic1,547±618,332—
17—muse-spark-1.1ClosedMeta1,542±87,549—
18—claude-opus-4-8ClosedAnthropic1,539±612,654—
19—claude-opus-4-6ClosedAnthropic1,537±519,536—
20—gemini-3.6-flash-highClosedGoogle1,537±88,259—
21—claude-sonnet-5-highClosedAnthropic1,537±79,296—
22—muse-spark-1.2 (xHigh)ClosedMeta1,534±142,065—
23—gpt-5.6-terra-xhigh (codex-harness)ClosedOpenAI1,521±87,694—
24—claude-sonnet-4-6ClosedAnthropic1,521±521,642—
25—gpt-5.6-luna-xhigh (codex-harness)ClosedOpenAI1,519±87,828—
26—gpt-5.5-xhigh (codex-harness)ClosedOpenAI1,510±614,946—
27—gemini-3.5-flash-highClosedGoogle1,500±78,931—
28—claude-opus-4-5-20251101-high-32kClosedAnthropic1,495±810,592—
29—gemini-3.5-flash-mediumClosedGoogle1,492±711,412—
30—gpt-5.5-high (codex-harness)ClosedOpenAI1,487±617,020—
31—claude-opus-4-5-20251101ClosedAnthropic1,468±712,899—
32—gpt-5.4-high (codex-harness)ClosedOpenAI1,463±191,277—
33—gpt-5.5 (codex-harness)ClosedOpenAI1,458±615,012—
34—gemini-3.1-pro-previewClosedGoogle1,447±522,891—
35—gemini-3.5-flash-liteClosedGoogle1,447±42228$0.85
36—gpt-5.4-medium (codex-harness)ClosedOpenAI1,443±191,316—
37—gemini-3-proClosedGoogle1,439±913,573—
38—gemini-3-flashClosedGoogle · 1 more variant1,438±910,453—
39—gpt-5-mediumClosedOpenAI1,420±163,032—
40—gpt-5.2ClosedOpenAI1,417±231,055—
41—inklingOpenThinking Machines1,410±78,970$1.8
42—gpt-5.3-codex (codex-harness)ClosedOpenAI · 1 more variant1,409±142,514—
43—Inkling SmallOpenThinking Machines1,407±95,058$0.52
44—gpt-5.4-mini-highClosedOpenAI1,397±711,279—
45—claude-sonnet-4-5-20250929-high-32kClosedAnthropic1,393±812,764—
46—gpt-5.1-mediumClosedOpenAI1,392±124,751—
47—claude-opus-4-1-20250805ClosedAnthropic1,389±116,769—
48—gpt-5.4ClosedOpenAI1,387±132,493—
49—claude-sonnet-4-5-20250929ClosedAnthropic1,385±715,150—
50—grok-4.20-beta-0309-reasoningClosedxAI1,374±614,571—

Intelligence over time

Each maker's best model at the time on the Artificial Analysis Intelligence Index, by the month it was released.

Gained most · Anthropic +57.2Highest now · Anthropic 64.6Gap between highest and lowest 18.1 ptsTime range Nov 2025 – Nov 2026
64.6Anthropic60.0Google56.9OpenAI50.1Meta46.5Kimi

Arena data comes from the leaderboard dataset it publishes on Hugging Face (CC BY 4.0). Only each board's overall leaderboard is used; scores and ranks are unchanged. dataset · CC BY 4.0

Artificial Analysis data comes from its API (artificialanalysis.ai). Scores are unchanged; ranks follow the scores. artificialanalysis.ai

Open or closed follows the license Arena gives; prices and Artificial Analysis intelligence show only where model names match, otherwise “—”.