The Global AI Market Intelligence Platform
See what's hot.
Know what's changing.
Act before the market moves.
Track AI model prices, token costs, inference trends, market heat and performance signals across the global AI economy.
hoton://newswire
LIVEHOTN-GTI
6.5 $/1M
Cheapest
$0.01
Models
537
- OpenAI Creates a New Framework to Disclose Bad AI Behavior2 hours ago
- Apple reportedly building server packed with M-series Ultra chips for AI2 hours ago
- Stanford Researchers Release Paper2Agent: Turning Research Papers Into AI Agents That Reproduce Results and Run on New Data2 hours ago
- Knowledgator Releases GLiFormer: A 575M-Parameter Encoder That Hits 91.10 F1 on Nested JSON Extraction Without Generating Tokens3 hours ago
- Anthropic and OpenAI want to embed safety evaluators. Will they really be independent?3 hours ago
- Washington Won’t Be Regulating AI Anytime Soon3 hours ago
- The 2.5-hour AI-generated Odyssey movie is 2.5 hours too long3 hours ago
- The AI data center e-waste problem is huge — and getting bigger3 hours ago
- After accusations of selling ‘perv glasses,’ Meta prepares to sell a pair without a camera4 hours ago
NewsView all →
Models Tracked
537
Providers Covered
79
Global Token Price Index
6.47
Model Efficiency Index
88.2
Live marketModels537Providers79Avg price$6.47/1MCheapest$0.01/1MMultimodal183Token Price Index6.47Data updated Sep 17, 2026
Latest activityChangelog →
00Why HotON
AI is becoming a global market.
HotON.ai helps you read it.
Models are changing. Prices are moving. Compute costs are shifting. HotON.ai turns fragmented AI signals into structured market intelligence.
01The AI Market Dashboard
Today's view of the global AI market
From model prices and usage trends to availability, latency and market heat — the whole AI economy on one screen.
Top Rated
Arena EloModelProviderEloOut / 1M
Cheapest Models
Blended $/1MLargest Context
K tokensMost Multimodal
ModalitiesTop Providers
ModelsCategory Distribution
Share of catalogText
58%
Multimodal
29%
Reasoning
8%
Code
5%
02Model Prices
Track the price of intelligence
AI pricing is no longer simple. HotON.ai helps builders, enterprises and investors understand how model costs move across providers and categories.
Input Token PriceOutput Token PriceContext Window CostMultimodal PricingEnterprise API Pricing
Model Price Board
$ / 1M tokens| Model | Region | Input | Output | Ctx |
|---|---|---|---|---|
| MINIMAX-M2.7 MiniMax | CN | $0.30 | $1.20 | 1000K |
| DEEPSEEK Novita AI | US | $0.14 | $0.28 | 1049K |
| QWEN3-CODER-FL Alibaba Cloud · Qwen | CN | $0.20 | $0.98 | 1000K |
| GEMINI-2.5-FLA Google | US | $0.10 | $0.40 | 1049K |
| GPT-4.1-MINI-2 OpenAI | US | $0.40 | $1.60 | 1048K |
| QWEN-PLUS Alibaba Cloud · Qwen | CN | $0.26 | $0.78 | 1000K |
| MINIMAX-01 MiniMax | CN | $0.20 | $1.10 | 1000K |
| GEMINI-3.1-FLA Google | US | $0.25 | $1.50 | 1049K |
| QWEN-PLUS-2025 Alibaba Cloud · Qwen | CN | $0.26 | $0.78 | 1000K |
| LLAMA-4-MAVERI Meta | US | $0.15 | $0.60 | 1049K |
| GEMINI-2.5-FLA Google | US | $0.10 | $0.40 | 1049K |
| GROK-4-FAST xAI | US | $0.20 | $0.50 | 2000K |
| GROK-4-1-FAST- xAI | US | $0.20 | $0.50 | 2000K |
| MINIMAX-M2.1 MiniMax | CN | $0.29 | $0.95 | 1000K |
Token Cost Calculator
DEMOCost / request
$0.001
Est. cost / month
$792.00
Cheapest for this workload: BAAI — $15.60/mo (save $776.40)
03AI Market Indexes
AI indexes for a new computing economy
Structured benchmarks for AI model prices, efficiency, inference cost and market momentum. News gets copied — indexes don't.
04Efficiency Ranking
Price is only one part of the story.
Efficiency tells the truth.
The cheapest model is not always the most efficient. HotON.ai compares models by total task cost, success rate, speed, stability and output quality.
| # | Model | Cost / Task | Context | Arena Elo | Efficiency |
|---|---|---|---|---|---|
| 01 | MINIMAX-M2.7 MiniMax | $0.001 | 1000K | — | 96 |
| 02 | DEEPSEEK Novita AI | $0.000 | 1049K | — | 96 |
| 03 | QWEN3-CODER-FL Alibaba Cloud · Qwen | $0.001 | 1000K | — | 96 |
| 04 | GEMINI-2.5-FLA Google | $0.000 | 1049K | — | 96 |
| 05 | GPT-4.1-MINI-2 OpenAI | $0.002 | 1048K | — | 96 |
| 06 | QWEN-PLUS Alibaba Cloud · Qwen | $0.001 | 1000K | — | 96 |
| 07 | MINIMAX-01 MiniMax | $0.001 | 1000K | — | 96 |
Find the model that actually delivers the best result for the cost. See the full price-vs-intelligence frontier →
05Inference Cost Map
Where is AI cheaper to run?
AI costs are shaped by more than model prices. Region, compute supply, energy cost, latency and availability all matter.
Global Compute Network
Provider regionsIllustrative map of provider regions — not a live feed.
Regional Cost Signals
Compare inference cost patterns across global regions.
Compute Availability
Understand where AI infrastructure capacity is becoming more attractive.
Energy-Aware AI
Track how energy conditions may influence compute and inference pricing.
Time-Based Cost Windows
Discover when certain regions may become more cost-efficient for AI workloads.
HotON.ai helps the market understand the geography of AI cost.
06AI Market Radar
The signals that actually matter
Model launches, pricing changes, infrastructure shifts, policy updates, funding events and market movements — filtered from the noise.
▲Sep 17, 2026■Sep 17, 2026▲Sep 17, 2026▲Sep 17, 2026■Sep 17, 2026■Sep 17, 2026■Sep 17, 2026■Sep 17, 2026■Sep 17, 2026■Sep 17, 2026■Sep 17, 2026■Sep 17, 2026■Sep 17, 2026■Sep 17, 2026■Sep 17, 2026▲Sep 17, 2026
Model Launches· WIRED· 2 hours ago
OpenAI Creates a New Framework to Disclose Bad AI Behavior
Model Launches· Ars Technica· 2 hours ago
Apple reportedly building server packed with M-series Ultra chips for AI
Model Launches· MarkTechPost· 2 hours ago
Stanford Researchers Release Paper2Agent: Turning Research Papers Into AI Agents That Reproduce Results and Run on New Data
Model Launches· MarkTechPost· 3 hours ago
Knowledgator Releases GLiFormer: A 575M-Parameter Encoder That Hits 91.10 F1 on Nested JSON Extraction Without Generating Tokens
Model Launches· TechCrunch· 3 hours ago
Anthropic and OpenAI want to embed safety evaluators. Will they really be independent?
Policy· WIRED· 3 hours ago
Washington Won’t Be Regulating AI Anytime Soon
Model Launches· The Verge· 3 hours ago
The 2.5-hour AI-generated Odyssey movie is 2.5 hours too long
Infrastructure· The Verge· 3 hours ago
The AI data center e-waste problem is huge — and getting bigger
Model Launches· TechCrunch· 4 hours ago
After accusations of selling ‘perv glasses,’ Meta prepares to sell a pair without a camera
Policy· The Decoder· 5 hours ago
EU president warns AI agents "escaping their environment" are just a preview of what's coming
Open Source· AWS ML· 5 hours ago
Improving HCLS AI reasoning with open-source agent skills
Open Source· WIRED· 5 hours ago
I Trained a Fly’s Brain to Generate WIRED Story Ideas
Model Launches· Hacker News· 5 hours ago
A coffee shop owner used AI to make a menu poster. Then came the angry DMs
Infrastructure· AWS ML· 5 hours ago
Fault tolerant distributed training on Amazon EKS using NVRx
Model Launches· TechCrunch· 6 hours ago
AI labs want in-house auditors — but maybe they should shut the front door first
Model Launches· The Decoder· 6 hours ago
Apple is reportedly building an enterprise AI server with its own M8 Ultra chips
HotON.ai Radar filters noise from the AI market and highlights the changes that may affect cost, access, capability and competition.
View all →07Market Media
Video, visuals and briefings — every format, one feed
HotON.ai delivers market intelligence as video, visuals and text. Choose how you consume the AI economy — the platform handles every format.
hoton://media
01:32
Video
Weekly AI Market Pulse
A 90-second video recap of the week's biggest moves in AI prices, models and infrastructure.
Watch recap
VisualInfographic
Global token price heatmap
Where input and output costs are rising and falling, at a glance.
Text Brief
Inference costs fall as new capacity comes online
Regional compute supply loosened this week, pushing the Inference Cost Index to a new monthly low across three major regions…
Video reports
Showreels, recaps and explainers with adaptive playback.
Visual & images
Charts, infographics and covers, responsive and crisp.
Text & briefings
Structured articles, summaries and data notes.
08HotON Reports
Reports built for AI decision makers
Structured intelligence on model pricing, AI infrastructure, inference cost, market heat and global AI supply-chain trends.
Briefings
Subscribe to HotON Reports
AI market briefings, pricing reports and index updates — in your inbox.
09HotON Data API
AI market data for builders and institutions
Access structured AI market data through HotON.ai APIs, feeds and custom intelligence products.
Model pricing feedsToken cost dataModel metadataIndex dataMarket heat rankingsLatency & availabilityReport feedsCustom intelligence
- Developers building AI tools
- Enterprises optimizing AI costs
- Model providers tracking position
- Investors following AI infrastructure
- Analysts researching the AI economy
api.hoton.ai
> GET /v1/models/OPUS-4.8/price { "symbol": "OPUS-4.8", "provider": "Anthropic", "input_per_1m": 6.00, "output_per_1m": 22.50, "context_k": 500, "efficiency": 96, "arena_elo": 1432, "modalities": ["text", "image"] }
Frequently asked questions
What is HotON.ai?+
HotON.ai is an AI market-intelligence platform tracking current prices (updated daily), token costs, quality (Arena Elo) and version price trends for 537 AI models across 79 providers worldwide.
Where does the pricing data come from, and how often is it updated?+
Prices come from each provider's official pricing (via the TestKey catalog), cross-checked against OpenRouter, and refreshed daily. Every model page shows the price's source and 'as of' date.
Is the data real?+
Yes. Model prices, context windows, modalities and Arena Elo scores are real and sourced; efficiency and cost-per-task are computed from those real inputs and labeled as such.
What is the cheapest AI model right now?+
As of Sep 17, 2026, the lowest blended price we track is $0.01 per 1M tokens. See the full 'cheapest models' ranking for the current list.
Which AI model is rated highest?+
By LMArena human-preference Elo, the top-rated model we track is claude-opus-4.6 (1505). See the quality ranking for the full leaderboard.
How is the efficiency / value score calculated?+
It is a composite of a model's real price and context window, normalized to 0-100 so cheaper, larger-context models score higher. Full details are on our Methodology page.
HotON.ai
Understand the AI market before everyone else.
HotON.ai gives you the data, indexes and intelligence to understand where the AI economy is moving next.