Ray 3
Luma AI • API Price Comparison
🏟️ Agent Arena: 1,205.4%Added Sep 15, 2026
Ray 3 is a large language model by Luma AI. Its Agent Arena net-improvement score is approximately 1205.4%. Prices are collected daily from official provider pages and every tracked channel, then cross-checked by a data-accuracy audit.
Model performance benchmarks
Scores retain their original benchmark, version and metric. We do not combine them into a site-wide score; results from different versions are not directly comparable.
1,205
Text-to-video leaderboardleaderboard-2026-09-15-versioned-slugs · ARENA_ELOAs of 2026-09-15
#31
Text-to-video leaderboardleaderboard-2026-09-15-versioned-slugs · ARENA_RANKAs of 2026-09-15
1,129
Text-to-video leaderboardleaderboard-2026-09-15-versioned-slugs · VOTESAs of 2026-09-15
💰 Cheapest Option
N/A
-/1M input
-/1M output
🏢 Official Price
N/A
No official channel available
🇨🇳 Best in China
No China-friendly options
📊 Channel Price Comparison
Compare ray-3 API prices across all available channels • 0 channels available
| Channel | Type | Input / 1M | Output / 1M | Rate Limit | China | vs Official |
|---|
📈 Price History
Every scraper-detected price change is recorded. Same-day multiple changes take the last value; unchanged channels carry the last observed price forward.
No price history yet. New price changes will appear after the next scraper run.
Based on typical usage patterns (input:output = 2:1 ratio); all channels converted to USD for comparison
| Usage Level | Tokens/Month |
|---|---|
| 🐣 Light | 0.1M in + 0.1M out |
| 📊 Medium | 1.0M in + 0.5M out |
| 🚀 Heavy | 10.0M in + 5.0M out |
| 🏢 Enterprise | 100.0M in + 50.0M out |
Frequently asked questions about ray-3
How often are Ray 3 prices updated, and how reliable is the data?
Prices are scraped daily from official pricing pages and every tracked channel, with every significant change recorded in a price-history log. A daily read-only data audit cross-checks each channel against the model producer's published price and flags stale or inconsistent rows for correction.