How the Price Index works
In plain English, then in detail.
In plain English
1. We pick 20 popular AI models — the ones most developers actually use (GPT, Claude, Gemini, Grok, DeepSeek, etc.).
2. Every day we check what each one costs on Qubax and what it costs on OpenRouter.
3. We combine all 20 ratios into one number. 100 means Qubax costs the same as OpenRouter. Below 100 means cheaper.
4. If the number goes down, Qubax got cheaper vs OpenRouter. If it goes up, the gap narrowed.
5. The lower the number, the more you save by using Qubax instead of OpenRouter.
What is the QPI?
The Qubax Price Index (QPI) is a daily parity index: it shows Qubax's blended price as a percentage of OpenRouter's price for the same basket of 20 flagship AI models.
100 = parity — Qubax costs exactly what OpenRouter charges. A value below 100 means Qubax is cheaper (e.g. 40 = 60% cheaper); above 100 means pricier. Because every day's value is computed directly against same-day OpenRouter prices, the index needs no launch-date calibration and can't be distorted by historical price glitches.
The Basket
The QPI basket contains 20 flagship models selected for:
- Brand recognition and developer search volume
- Coverage across major AI labs (OpenAI, Anthropic, Google, xAI, DeepSeek, Meta, Alibaba, Moonshot, Zhipu)
- At least 25 days of continuous price history
- Diversity across model tiers (flagship, fast, small, coding)
The basket is fixed — the same models are tracked every day. This ensures the index measures genuine price movement, not changes in what models are available.
Price Formula
Blended Price
blended_price = (input_price × 3 + output_price × 1) / 4Uses a 3:1 input:output ratio, reflecting typical chat and coding workloads where input tokens dominate. Prices are per 1 million tokens, in USD.
Index Value
QPI = median(qubax_blended_price / openrouter_blended_price) × 100The median (not mean) is used to prevent a single model's price spike from dominating the index. Each day's ratio is computed against same-day OpenRouter prices — no baseline window, no calibration date. Ratios are winsorized to [0.01, 5.0] so a single model crash/spike can't dominate.
OpenRouter Benchmark Line
The gray dashed line on the chart is the flat 100 parity reference. The QPI line's distance below it is your savings: at 40, the median basket model costs 60% less on Qubax than on OpenRouter. A separate self-indexed trend of OpenRouter's own prices is retained in the data for reference but the parity gap is the headline.
Data Provenance
- Qubax prices: live sell prices from the Qubax platform, updated daily at 08:00 UTC
- Benchmark prices: OpenRouter public models API (
openrouter.ai/api/v1/models) - Historical data: 75,000+ price snapshots dating to June 24, 2026
- Computation: server-side, stored in
price_index_snapshotstable
Limitations
- The basket covers text models only; image, video, and audio models are excluded
- Models that are delisted or unavailable for >7 consecutive days are temporarily excluded
- The 3:1 blend ratio is a simplification; actual usage patterns vary by application
- The index measures list prices, not negotiated or volume-discounted rates