Skip to main content
BenchLM
Data
Specialist vertical · BenchAlign v6.0 · 76 ranked (38 Supported, 38 Estimated)
Scored from the dataset of

Frontend and web development leaderboard

As of November 21, 2026, Claude Fable 5.2 leads our frontend and web development leaderboard at 89.5 (Supported, $8/$40 per 1M tokens), ahead of Claude Opus 5.5 at 82.8. GPT-6.1 Sol (#3, 81.2) is the lowest-cost model near the top at $2/$10 per 1M tokens; Kimi K3.1 Pro leads open-weight models at #15 (70.0). We rank 76 models (38 Supported, 38 Estimated) across 9 domain protocols, and this vertical carries zero weight in the overall score.

Shortlist by workload and budget

Each pick comes directly from the live frontend & web dev rows below, with listed API rates and the case where another option wins.

Highest point score#1 · Supported

Anthropic · $8/$40 per 1M tokens

Costs $8/$40 per 1M tokens; GPT-6.1 Sol (#3, 81.2) cuts token spend to $2/$10 per 1M tokens.

Closest contender#2 · Supported

Anthropic · $4/$20 per 1M tokens

Sits 6.7 points behind Claude Fable 5.2; test both on your own prompts when provider lock-in or latency matters.

Best API value#3 · Supported

OpenAI · $2/$10 per 1M tokens

Keeps 91% of the top score (81.2 vs 89.5), but gives up 8.3 points on the hardest tail cases.

Top open-weight model#15 · Supported

Moonshot AI · Open weights · $2/$10 per 1M tokens

Trails Claude Fable 5.2 by 19.5 points (#15 overall here), so pick it when weight control or on-prem residency outweighs peak score.

Ranked models

Ordered by BenchAlign v6.0 point estimate as of November 21, 2026. An Estimated row backed by fewer than two source families never ranks above a row scoring within 2 points below it. Conditional ranges do not establish rank confidence.

Frontend & WebDev LLM Leaderboard
RankModelScoreEvidence
#1Anthropic89.5Supported
#2Anthropic82.8Supported
#3OpenAI81.2Supported
#4OpenAI80.2Supported
#5Anthropic77.6Supported
#6Anthropic78.0Estimated
#7Google77.9Estimated
#8OpenAI75.8Supported
#9OpenAI75.8Supported
#10Anthropic74.3Supported
#11OpenAI74.0Supported
#12Alibaba74.0Estimated
#13Google71.9Supported
#14Anthropic70.7Supported
#15
Kimi K3.1 ProOpen weights
Moonshot AI
70.0Supported
#16Google71.6Estimated
#17DeepSeek69.2Supported
#18xAI69.1Supported
#19OpenAI66.5Supported
#20Anthropic66.2Estimated
#21OpenAI65.9Supported
#22
Qwen3.8 MaxOpen weights
Alibaba
65.8Supported
#23OpenAI64.9Supported
#24OpenAI66.7Estimated
#25Google64.6Supported
#26
Kimi K3Open weights
Moonshot AI
64.6Supported
#27MiniMax64.3Supported
#28Google64.3Supported
#29OpenAI63.9Supported
#30
MiMo-V2.6-ProOpen weights
Xiaomi
65.7Estimated
#31Anthropic63.6Supported
#32
GLM-5.3Open weights
Z.AI
63.5Supported
#33Meta65.4Estimated
#34
GLM-5.3-FlashOpen weights
Z.AI
63.2Supported
#35Google64.1Estimated
#36
Qwen3.8-27BOpen weights
Alibaba
62.4Estimated
#37Google60.2Supported
#38xAI62.1Estimated
#39
Step 5 PreviewOpen weights
StepFun
61.7Estimated
#40
GLM-5.2Open weights
Z.AI
58.8Supported
#41OpenAI58.4Supported
#42Google60.4Estimated
#43
DeepSeek V4 ProOpen weights
DeepSeek
60.0Estimated
#44Meta57.8Supported
#45Anthropic57.4Supported
#46Anthropic57.2Supported
#47Anthropic57.2Supported
#48Anthropic57.2Supported
#49Alibaba58.3Estimated
#50Meta56.8Estimated
#51Google54.0Supported
#52
GLM-5.1Open weights
Z.AI
54.5Estimated
#53Anthropic52.3Estimated
#54Anthropic52.3Estimated
#55
MiniMax M3Open weights
MiniMax
52.1Supported
#56Google53.8Estimated
#57Anthropic51.9Estimated
#58OpenAI51.4Estimated
#59OpenAI49.5Estimated
#60
Mistral Large 4Open weights
Mistral
48.9Estimated
#61
Qwen3.5 397BOpen weights
Alibaba
46.0Estimated
#62Anthropic46.2Estimated
#63Anthropic46.2Estimated
#64
Muse Spark OpenOpen weights
Meta
44.2Estimated
#65xAI43.1Estimated
#66
Kimi K2.5Open weights
Moonshot AI
41.1Supported
#67
Gemma 4 31BOpen weights
Google
40.6Supported
#68
GLM-5Open weights
Z.AI
41.6Estimated
#69Google40.4Estimated
#70
InklingOpen weights
Thinking Machines Lab
37.5Estimated
#71
MiniMax M2.7Open weights
MiniMax
36.2Estimated
#72
DeepSeek V3.2Open weights
DeepSeek
34.7Estimated
#73
GPT-OSS 120BOpen weights
OpenAI
33.8Estimated
#74Anthropic31.9Estimated
#75Anthropic31.9Estimated
#76
GPT-OSS 20BOpen weights
OpenAI
29.8Estimated

Scored domain benchmarks (9)

Each percentage is the protocol's relative reference weight inside this vertical's calibrated BenchAlign v6.0 model, not a raw score average. Independent evaluator shares are capped at 40% whenever three or more independent evaluators contribute.

A GUI-grounding benchmark for 1,581 instructions in full-screen, high-resolution professional interfaces. It tests where a target is, not whether an agent can finish the surrounding workflow.

17% ref. weight

A live human-preference benchmark where language models turn text prompts into voxel structures and voters compare anonymous builds from the same prompt.

16% ref. weight

Contamination-resistant frontend interface and interactive component design tasks.

16% ref. weight

Evaluates coding agents on 333 multimodal game-development tasks in Godot, spanning 2D graphics, 3D graphics, user interfaces, and gameplay logic.

14% ref. weight

A coding-agent benchmark for realistic React work, with rubrics that check production concerns such as performance, accessibility, correctness, and code quality.

11% ref. weight

HeyGen's Code2Video tests whether a model can turn a detailed creative brief into a finished motion-graphics video by writing code.

11% ref. weight

Coding agents recreate nine reference web apps from an empty directory. Plato checks visual and functional behavior through end-to-end workflows.

8% ref. weight

An open benchmark for AI coding agents on real-world React Native implementation tasks, emphasizing working app behavior, recommended architecture choices, and strict constraint adherence.

7% ref. weight

A Vercel benchmark for AI coding agents on Next.js code generation and migration tasks, reporting pass@4, average execution time, and results with bundled documentation through AGENTS.md.

2% ref. weight

Executable React components, visual layout cloning, and pixel-level UI grounding

Nine benchmarks shape this vertical, led by ScreenSpot Pro (17%), VoxelBench Text (16%), LiveBench 2.0 Design (16%) alongside GameDevBench, ReactBench, Code2Video, Web Clone Bench, React Native Evals, and Next.js Evals. Rather than grading isolated Python algorithms, these suites execute real component trees, check interactive state updates in React and Next.js, ground cursor coordinates on high-density desktop viewports, and render 2D/3D browser scenes.

Thirty-eight of the 76 models hold Supported status because multiple independent evaluators—including ScreenSpot Pro, LiveBench, and framework-specific boards—test the same public checkpoints.

Passing component assertions does not guarantee tasteful UI or accessible markup

Automated DOM harnesses and visual-similarity metrics reward layout match and state transitions, but they rarely catch generic SaaS card grids, awkward mobile tap targets, missing ARIA focus traps, or bloated client bundles. A model can ace ReactBench while still importing deprecated Next.js router patterns unless your prompt supplies current framework docs.

Keep Lighthouse, axe-core accessibility checks, and bundle-size budgets in continuous integration regardless of which model writes your JSX.

When to pay for the frontier leader versus fast mid-tier iteration

Claude Fable 5.2 (89.5, Supported, $8/$40 per 1M tokens) holds a 6.7-point lead over Claude Opus 5.5 (82.8, Supported, $4/$20 per 1M tokens), one of the widest #1-to-#2 gaps on any BenchAlign v6.0 board. That margin shows up on complex multi-file React refactors and interactive canvas tasks where a single broken hook stalls the preview.

For high-frequency component scaffolding or design-system variants where $8/$40 tokens add up quickly, GPT-6.1 Sol (#3, 81.2, Supported at $2/$10 per 1M tokens) cuts API spend by 75% while staying above 81 points, and Kimi K3.1 Pro (#15, 70.0, Supported) leads self-hosted weights.

Questions

Which LLM is best for frontend development right now?

Claude Fable 5.2 holds #1 on our frontend development leaderboard at 89.5 (Supported) as of November 21, 2026, ahead of Claude Opus 5.5 at 82.8 and GPT-6.1 Sol at 81.2. Both top models hold Supported status across independent public evaluations.

What is the best open-weight LLM for frontend development?

Kimi K3.1 Pro from Moonshot AI leads open-weight models on this frontend development board at #15 with a calibrated score of 70.0 (Supported). That trails Claude Fable 5.2 by 19.5 points, so choose Kimi K3.1 Pro when self-hosting, VPC data residency, or weight customization matters more than peak frontier accuracy.

Which API model offers the best value for frontend development?

GPT-6.1 Sol (#3, 81.2) is our top API value pick for frontend development, priced at $2/$10 per 1M tokens compared with Claude Fable 5.2 at $8/$40 per 1M tokens. You give up 8.3 calibrated points, making GPT-6.1 Sol the better buy whenever those extra points do not cut downstream retries.

Which benchmarks make up the frontend development leaderboard score?

BenchAlign v6.0 calibrates 9 open and primary domain protocols for this board, led by ScreenSpot Pro (17%), VoxelBench Text (16%), LiveBench 2.0 Design (16%), GameDevBench (14%). Each percentage reflects share of the vertical's reference weight inside our item-response model rather than a raw score average, and no single independent evaluator can exceed 40% when three or more evaluators contribute.

Does the frontend leaderboard test real React and Next.js execution?

Yes. ReactBench, Next.js Evals, React Native Evals, Web Clone Bench, and GameDevBench execute generated component code and browser apps against functional assertions and visual layout targets, while ScreenSpot Pro tests high-resolution UI coordinate grounding. Static syntax fluency alone cannot earn a top frontend score here.

What frontend qualities are missing from automated web-dev benchmarks?

Automated UI benchmarks verify DOM state, layout geometry, and task completion, but they do not grade brand typography, keyboard focus management, screen-reader WCAG compliance, Core Web Vitals bundle weight, or API security headers. Keep automated accessibility and bundle audits in CI alongside your chosen coding model.

Frontend & web dev leaderboard updates

Weekly updates when new models or domain evaluations move the frontend & web dev rankings.

Read a sample issue

Join 7,000+ subscribers.

One email each week. Unsubscribe anytime.