What is the LLM Leaderboard?

The LLM Leaderboard by Teh Tarik Digital is a real-time tracking tool that compares hundreds of Large Language Model (LLM) listings from providers like OpenAI, Anthropic, Google, Meta, and DeepSeek based on intelligence (Arena Score), per-model capability breakdowns, context length, latency mode, and API cost per million tokens.

How are the LLM scores calculated?

Overall intelligence ratings use Elo-based Arena Scores sourced from LMSYS Chatbot Arena. Per-model capability breakdowns (reasoning, math, coding, and dozens of other skill categories) and named benchmark leaderboards — including SWE-Bench Verified, LiveCodeBench, GPQA, AIME 2025, MMLU-Pro, MMMU-Pro, MMLU, IFEval, Humanity's Last Exam, Terminal-Bench 2.0, and Finance Agent — are integrated from llm-stats.com. Named benchmarks are scored independently of the Arena ranking, so a model can lead a specific benchmark without ranking highly overall.

How does LLM API pricing work on this leaderboard?

Pricing is listed per 1 million input and output tokens across Standard real-time API, Batch processing (50% off), and Fast/Priority queues. Prices can be toggled between USD ($) and Malaysian Ringgit (MYR RM) using live foreign exchange rates.

LLM Leaderboard - Compare AI Models, Benchmarks & Pricing

Q: Why does the leaderboard show more listings than distinct models?

Each row is a distinct listing from OpenRouter's catalog, not necessarily a distinct model. The same underlying model is frequently listed more than once under different execution modes — Standard, Batch, and Fast/Priority — each with its own row, pricing, and sometimes its own Arena rank.

A live, continuously-updated registry of AI model listings from every major lab — filterable by coding, vision, reasoning, long context, and price-to-performance, with per-model capability breakdowns, execution-mode pricing (standard, batch, fast), and real API cost in USD or MYR. Built to answer one question: what's actually worth building on right now.

UPDATED - -- -- USD/MYR -.--- ARENA -- · CODING --

Top ranked

Claude Fable 5

Anthropic

1,507

Arena Score

$10.00

Input / 1M

Execution Modes & Pricing Breakdown

StandardReal-Time API

Synchronous requests with standard latency and baseline pricing. Ideal for live production apps and chatbots.

Batch (-50%)Async Off-Peak

24-hour completion window at a 50% discount using off-peak provider GPU compute. Ideal for evaluations and bulk workloads.

Fast / PriorityLow Latency

Priority queueing and reserved throughput for high concurrency or ultra-fast response times. Premium pricing.

Loading models...

Price vs. Arena Score

Lower-right = best value

No scored models match the current filters — try a different tab or clear filters.

Scale

Warp 1.00

Bandwidth 0.12

Density curve

Model	Provider	Arena Score	Context	Input / 1M	Output / 1M
Claude Fable 5	Anthropic	1,507	1000k	$10.00	$50.00
Claude Fable 5	Anthropic	1,507	1000k	$5.00	$25.00
Muse Spark 1.2	Meta	1,499	1049k	$1.25	$4.25
Claude Opus 4.6	Anthropic	1,498	1000k	$5.00	$25.00
Claude Opus 4.6	Anthropic	1,498	1000k	$2.50	$12.50
Claude Opus 4.7	Anthropic	1,494	1000k	$5.00	$25.00
Claude Opus 4.7	Anthropic	1,494	1000k	$2.50	$12.50
Muse Spark 1.1	Meta	1,492	1049k	$1.25	$4.25
Gemini 3.1 Pro Preview	Google	1,487	1049k	$2.00	$12.00
Gemini 3.1 Pro Preview	Google	1,487	1049k	$1.00	$6.00
Qwen3.8 Max	Qwen	1,479	1000k	$2.00	$6.00
GPT-5.5	Openai	1,477	1050k	$5.00	$30.00
GPT-5.5	Openai	1,477	1050k	$2.50	$15.00
GLM 5.3 Flash	Z-ai	1,474	1311k	$0.08	$0.25
GLM 5.3 Flash	Z-ai	1,474	1049k	$0.15	$0.50
Claude Opus 4.8	Anthropic	1,473	1000k	$5.00	$25.00
Claude Opus 4.8	Anthropic	1,473	1000k	$2.50	$12.50
Claude Sonnet 4.6	Anthropic	1,472	1000k	$3.00	$15.00
Claude Sonnet 4.6	Anthropic	1,472	1000k	$1.50	$7.50
Grok 4.5	X-ai	1,470	500k	$2.00	$6.00
MiMo-V2.5-Pro	Xiaomi	1,468	1050k	$0.44	$0.87
GLM 5.1	Z-ai	1,466	205k	$0.97	$3.04
GPT-5.4	Openai	1,466	1050k	$2.50	$15.00
GPT-5.4	Openai	1,466	1050k	$1.25	$7.50
Kimi K2.6	Moonshotai	1,461	262k	$0.95	$4.00

Rows per page:

Page 1 of 1

About the model count: Each row here is a distinct listing from OpenRouter's catalog, not necessarily a distinct model — the same underlying model is frequently listed more than once under different execution modes (Standard, Batch, Fast/Priority — see the ), each with its own row, pricing, and sometimes its own Arena rank. So "300+ models tracked" means 300+ listings, not 300+ distinct labs' worth of genuinely different models.

About the chart above: The price-vs-score chart only plots models with a real, matched Arena score (or, on the Coding tab, a matched coding benchmark score) — a model can be listed in the table below with full pricing and context info while having no score yet, either because it's new, hasn't been ranked on Arena, or didn't match cleanly against Arena's naming. Those models are still tracked and visible in the table; they just don't have a point on the chart.

INDEPENDENT SCORING · LLM-STATS.COM · OPENROUTER

Benchmarks & Model Insights

Real-time speed, latency, and pricing rankings from OpenRouter, a side-by-side model comparison tool, and named benchmark leaderboards — SWE-Bench, GPQA, MMLU-Pro, and others — scored independently of the Arena ranking above. A model can lead here without appearing in the main table's top ranks, since these measure specific skills and real infrastructure performance rather than general preference.

Fastest Models

(ranked by OpenRouter)

Based on median response time. Lower is better.

Lowest Latency

(ranked by OpenRouter)

Median time to first token. Lower is better.

Cheapest Models

(per 1M tokens)

View full ranking

$ Input / $ Output per 1M tokens. Lower is better.

Fastest and Lowest Latency rank across OpenRouter's full model catalog — including smaller and specialty providers — not just the major labs shown elsewhere on this page.

Compare Models

Side-by-side performance and benchmark comparison

Performance metrics Real-time benchmarks Side-by-side insights

Pick two models to compare performance across benchmarks.

–

Models Tracked

Continuously monitored across providers

Real-time Data

Live benchmarks and performance metrics

Open & Transparent

Data from OpenRouter and industry sources

Updated Daily

Fresh data every day so you stay ahead