Pricing Trends Tracker
Analyzes how model pricing distributes across 300 AI models and which providers offer the best deals. Compare input vs output costs, price tiers, and find the highest-scoring model at every budget.
Pricing Overview
Price vs Score (Paid Models)
Models per Price Tier
Price Tier Distribution
| Tier | Count | Avg Score |
|---|---|---|
| Free | 17 | 45 |
| Ultra-Budget | 62 | 59 |
| Budget | 89 | 67 |
| Mid-Range | 92 | 74 |
| Premium | 20 | 74 |
| Enterprise | 20 | 84 |
Full Pricing Table
Top 80 models by scoreInput vs Output Pricing Analysis
I/O Ratio = output price / input price. A ratio of 3.0x means output tokens cost 3x more than input tokens.
Highest I/O Ratio
Most expensive output relative to input| Model | $/M In | $/M Out | Ratio |
|---|---|---|---|
| Qwen3 Next 80B A3B Instruct | $0.09 | $1.10 | 12.2x |
| Qwen3 30B A3B Thinking 2507 | $0.20 | $2.40 | 12.0x |
| Perceptron Mk1 | $0.15 | $1.50 | 10.0x |
| Palmyra X5 | $0.60 | $6.00 | 10.0x |
| Qwen3 VL 235B A22B Thinking | $0.40 | $4.00 | 10.0x |
| Qwen3 235B A22B Thinking 2507 | $0.23 | $2.30 | 10.0x |
| Qwen3 VL 235B A22B Instruct | $0.21 | $1.90 | 9.0x |
| Gemini 3.5 Flash Lite | $0.30 | $2.50 | 8.3x |
| Gemini 3.5 Flash Lite (batch) | $0.15 | $1.25 | 8.3x |
| Ring-2.6-1T | $0.07 | $0.63 | 8.3x |
Lowest I/O Ratio
Most balanced pricing| Model | $/M In | $/M Out | Ratio |
|---|---|---|---|
| Reka Edge | $0.10 | $0.10 | 1.0x |
| R1 Distill Llama 70B | $0.80 | $0.80 | 1.0x |
| Llama 3.1 70B Instruct | $0.40 | $0.40 | 1.0x |
| Gemma 2 27B | $0.65 | $0.65 | 1.0x |
| DeepSeek V3.2 | $0.27 | $0.40 | 1.5x |
| Qwen3.5-9B | $0.10 | $0.15 | 1.5x |
| DeepSeek V3.2 Exp | $0.27 | $0.41 | 1.5x |
| Llama 3.1 8B Instruct | $0.05 | $0.08 | 1.6x |
| DeepSeek V4 Flash 0731 | $0.09 | $0.18 | 2.0x |
| Laguna S 2.1 | $0.09 | $0.18 | 2.0x |
Best Value at Each Price Point
The highest-scoring model at or below each average cost threshold.
| Max $/1M | Model | Score | Actual Cost |
|---|---|---|---|
| $0.10 | Gemma 4 31B (free) | 81 | Free |
| $0.50 | GPT-5.6 Luna Pro | 89 | $0.35/1M |
| $1.00 | GPT-5.6 Luna Pro | 89 | $0.35/1M |
| $2.00 | GPT-5.6 Luna Pro | 89 | $0.35/1M |
| $5.00 | Gemini 3.1 Pro Preview (batch) | 92 | $3.50/1M |
| $10.00 | Claude Opus 4.7 (batch) | 95 | $7.50/1M |
| $20.00 | Claude Fable 5 (batch) | 97 | $15.00/1M |
| $50.00 | Claude Fable 5 | 97 | $30.00/1M |
Provider Pricing Strategy
Providers with 2+ models, sorted by avg cost| Provider | Models | Avg Cost |
|---|---|---|
| TII | 3 | Free |
| poolside | 4 | $0.06 |
| inclusionai | 5 | $0.15 |
| Tencent | 2 | $0.23 |
| Meta | 5 | $0.27 |
| NVIDIA | 9 | $0.43 |
| Xiaomi | 2 | $0.43 |
| StepFun | 2 | $0.44 |
| DeepSeek | 12 | $0.66 |
| ByteDance | 4 | $0.67 |
| MiniMax | 8 | $0.74 |
| Kuaishou | 3 | $0.99 |
| Alibaba | 31 | $1.18 |
| Zhipu AI | 13 | $1.29 |
| thinkingmachines | 3 | $1.53 |
| xAI | 5 | $2.23 |
| aion-labs | 3 | $2.25 |
| 27 | $2.26 | |
| Moonshot AI | 8 | $2.51 |
| meta | 2 | $2.75 |
| Mistral AI | 6 | $2.98 |
| Cursor | 2 | $3.00 |
| Cohere | 4 | $3.22 |
| 2 | $5.75 | |
| ~openai | 2 | $10.06 |
| ~anthropic | 4 | $13.50 |
| Anthropic | 28 | $16.79 |
| OpenAI | 86 | $18.93 |
We categorize models into tiers based on average cost per million tokens: Free (zero cost), Budget (under $1), Mid ($1-$10), Premium ($10-$50), and Enterprise (over $50). This tiering helps developers quickly find models that match their budget constraints.
The I/O ratio compares the cost of output tokens to input tokens. A ratio of 3x means output costs three times more than input. This matters because output-heavy workloads (like content generation) will cost more with high-ratio models, while input-heavy workloads (like analysis) are less affected.
Use the "Best Value at Each Price Point" table to find the highest-scoring model within your budget. Set your maximum cost threshold, and the table shows which model delivers the best quality at or below that price. You can also check the scatter chart to visually identify models with the best score-to-cost ratio.