SIT Methodology

2.1Unit

1 SIT = 1 million tokens of inference at a defined quality standard.

"Tokens" refers to the standard industry unit of LLM inference, as measured by each provider's tokenizer. While tokenizers differ between models, the "million tokens" convention is universally adopted and provides sufficient standardization for pricing purposes.

2.2Pricing Components

Every SIT-eligible model has three published prices:

Blended price formula:

blended_price = (0.4 × input_price) + (0.6 × output_price)

The 60% output weighting reflects production workloads where output tokens exceed input tokens (generation, coding, summarization). This ratio will be refined as real-world usage data becomes available.

4.1Tier Indices

Each quality tier has its own composite index:

4.2Weighting Methodology

Phase 1 (Launch): Equal weighting across all models within each tier.

SIT-Composite = Σ(blended_price_i × weight_i) / Σ(weight_i)
where weight_i = 1.0 (equal weight)

Phase 2 (3–6 months): Capacity-weighted. Models weighted by context window and provider size.

Phase 3 (6–12 months): Volume-weighted. Models weighted by actual API transaction volume, sourced from provider-reported volumes and OpenRouter routing volumes.

4.3Calculation Frequency

4.4Base Date and Rebaselining

• Base date: August 3, 2026
• Base value: SIT-Composite = 1000 index points
• Rebaselining only on methodology changes. All rebaselining events are published with full explanation and a 14-day public comment period.

6.1Primary Sources

6.2Source Hierarchy

When a model is available from multiple sources, priority:

1. Direct provider (e.g. openai.com pricing for GPT-5.6)
2. Aggregator with lowest markup (e.g. OpenRouter base price)
3. Community submission (verified against at least one other source)

6.3Data Quality

• Every price point stores: timestamp, source URL, raw price, normalized price
• Automated anomaly detection: if a price moves >50% in one hour, flagged
• Manual review of all tier additions and removals
• Full audit trail: every index calculation is reproducible from stored raw data

7.1Methodology Changes

Any change to this methodology triggers:

1. 14-day public comment period
2. Full version increment (0.1 → 0.2)
3. Recalculation of historical indices using new methodology
4. Publication of both old and new values for 30-day overlap

7.2Conflict of Interest

• InferenceIndexer is an independent price reporting agency
• InferenceIndexer does not provide inference services
• InferenceIndexer does not take positions in inference futures or derivatives
• All data sources are public and verifiable
• Methodology is fully transparent and reproducible

8.1Known Limitations

1. Tokenizer differences: Different models use different tokenizers. A "million tokens" from GPT-5.6 processes more text than a million tokens from Llama 3.2. This is analogous to different crude oil grades having different energy densities. The SIT accepts this imprecision as the cost of standardization.
2. Volume data: Phase 1 uses equal weighting because real-world transaction volumes are not publicly available. The index may over-represent niche models.
3. Aggregator dependency: Many prices are sourced via OpenRouter. If OpenRouter changes its pricing model, coverage may temporarily decrease.
4. Quality benchmark dependency: Tier assignments depend on the Artificial Analysis Intelligence Index. Changes to their methodology affect our tiers.
5. Excluded models: Per-request pricing, enterprise-only pricing, and deprecated models are not tracked.

8.2Future Enhancements

• Volume weighting with real API call volumes
• Latency-adjusted pricing (tokens/second as a factor)
• Cache pricing tracked separately
• Batch pricing tracked separately
• Regional pricing (US, EU, Asia)