Frontier Basket USD/SCU
$6.4677
same live basket as homepage
Experimental / simulated
Estimated SCU data
Standard Capability Unit
SCU is not a hardware compute metric. It does not measure GPU FLOPs, training cost, or raw inference compute. It measures capability-adjusted token output.
TPI measures token price. TPX measures token purchasing power. mToken to SCU measures how much standardized AI capability one million standard workload tokens can deliver.
Current SCU values are public-data estimates. Estimated USD/SCU is a synthetic quote, not a traded market price. Experimental Live SCU uses simulated delivery factors until real provider telemetry is connected.
US Basket USD/SCU
$7.8175
qualified US model average
Experimental / simulated
China Basket USD/SCU
$1.4701
qualified China model average
Experimental / simulated
SCU Purchasing Power
0.1546 SCU
$1 at live basket quote
Experimental / simulated
What SCU Measures
Dollars buy tokens. Tokens generate outputs. SCU measures how much practical AI capability those outputs actually deliver.
TPI and TPX remain token-economic indices. SCU adds a capability layer on top of token economics: TPI is the price layer, TPX is the purchasing-power layer, and SCU is the capability-adjusted token value layer.
Why SCU Exists
USD is the payment unit, model tokens are the invocation unit, and SCU is the capability delivery unit. Different model tokens cannot be compared by price alone because quality, success rate, reliability, and task coverage differ.
Token-to-SCU Exchange Rates
Prototype rates reuse the current Sources model list and price snapshots, then attach estimated AI-native benchmark metadata, bounded quality adjustments, data coverage, and confidence.
OpenAI
GPT-5.5
- Region
- United States
- USD/1M
- 5.0000 / 30.0000
- mToken -> SCU
- 1.0318
- mToken -> F-SCU
- 1.032
- Est. USD/SCU
- 12.6604
- Coverage
- 83%
- Confidence
- Medium
- Source reliability
- official
- Last updated
- May 17, 2026, 12:00 AM
Anthropic
Claude Opus 4.7
- Region
- United States
- USD/1M
- 5.0000 / 25.0000
- mToken -> SCU
- 1.0066
- mToken -> F-SCU
- 1.007
- Est. USD/SCU
- 11.3753
- Coverage
- 82%
- Confidence
- Medium
- Source reliability
- official
- Last updated
- May 17, 2026, 12:00 AM
Gemini 3.1 Pro Preview
- Region
- United States
- USD/1M
- 2.0000 / 12.0000
- mToken -> SCU
- 0.9617
- mToken -> F-SCU
- 0.962
- Est. USD/SCU
- 5.4333
- Coverage
- 80%
- Confidence
- Medium
- Source reliability
- official
- Last updated
- May 17, 2026, 12:00 AM
xAI
Grok 4.3
- Region
- United States
- USD/1M
- 1.2500 / 2.5000
- mToken -> SCU
- 0.9178
- mToken -> F-SCU
- 0.918
- Est. USD/SCU
- 1.8012
- Coverage
- 76%
- Confidence
- Medium
- Source reliability
- official
- Last updated
- May 17, 2026, 12:00 AM
DeepSeek
DeepSeek V4 Pro
- Region
- China
- USD/1M
- 0.4350 / 0.8700
- mToken -> SCU
- 0.8865
- mToken -> F-SCU
- 0.887
- Est. USD/SCU
- 0.6489
- Coverage
- 78%
- Confidence
- Medium
- Source reliability
- official
- Last updated
- May 17, 2026, 12:00 AM
Alibaba Qwen
Qwen3 Max
- Region
- China
- USD/1M
- 0.3590 / 1.4340
- mToken -> SCU
- 0.8719
- mToken -> F-SCU
- 0.872
- Est. USD/SCU
- 0.8094
- Coverage
- 76%
- Confidence
- Medium
- Source reliability
- official
- Last updated
- May 17, 2026, 12:00 AM
Moonshot Kimi
Kimi K2.6
- Region
- China
- USD/1M
- 0.6000 / 2.5000
- mToken -> SCU
- 0.8512
- mToken -> F-SCU
- 0.851
- Est. USD/SCU
- 1.4248
- Coverage
- 68%
- Confidence
- Low
- Source reliability
- official
- Last updated
- May 17, 2026, 12:00 AM
ByteDance Doubao
Doubao Seed 1.8
- Region
- China
- USD/1M
- 0.8000 / 2.0000
- mToken -> SCU
- 0.8394
- mToken -> F-SCU
- 0.839
- Est. USD/SCU
- 1.4142
- Coverage
- 68%
- Confidence
- Low
- Source reliability
- official
- Last updated
- May 17, 2026, 12:00 AM
Baidu
ERNIE 5.0
- Region
- China
- USD/1M
- 1.7000 / 5.5000
- mToken -> SCU
- 0.8264
- mToken -> F-SCU
- 0.826
- Est. USD/SCU
- 3.5402
- Coverage
- 66%
- Confidence
- Low
- Source reliability
- official
- Last updated
- May 17, 2026, 12:00 AM
Zhipu
GLM-4.6
- Region
- China
- USD/1M
- 1.4000 / 1.4000
- mToken -> SCU
- 0.8134
- mToken -> F-SCU
- 0.813
- Est. USD/SCU
- 1.7212
- Coverage
- 66%
- Confidence
- Low
- Source reliability
- official
- Last updated
- May 17, 2026, 12:00 AM
MiniMax
MiniMax-M2.7
- Region
- China
- USD/1M
- 0.3000 / 1.2000
- mToken -> SCU
- 0.8062
- mToken -> F-SCU
- 0.806
- Est. USD/SCU
- 0.7321
- Coverage
- 73%
- Confidence
- Medium
- Source reliability
- official
- Last updated
- May 17, 2026, 12:00 AM
Benchmark Basket Weights
- reasoning
- 15%
- coding
- 20%
- agent
- 20%
- long Context
- 10%
- multimodal
- 10%
- verification
- 10%
- robustness
- 10%
- human Benchmark Bridge
- 5%
Standardized Workload Profiles
| Profile | Weight | Input | Output |
|---|---|---|---|
| General QA | 10% | 60% | 40% |
| Reasoning | 15% | 50% | 50% |
| Coding | 20% | 65% | 35% |
| Code repair | 10% | 80% | 20% |
| Long context | 10% | 90% | 10% |
| Multimodal | 10% | 70% | 30% |
| Agent / tool use | 15% | 75% | 25% |
| Verification | 10% | 60% | 40% |
SCU Formula
AI-native Capability
AI_Native_Capability_i,t = Sum_c category_weight_c x normalized_category_score_i,c,t
Estimated category scores are a placeholder for a public AI-native benchmark basket covering reasoning, code, agents, long context, multimodal work, verification, robustness, and human benchmark bridge metrics.
Quality Adjustment
QualityAdjustedCapability_i,t = AI_Native_Capability_i,t x ReliabilityAdjustment_i,t x VerificationAdjustment_i,t x DataConfidenceAdjustment_i,t
Reliability, verification, and data confidence use bounded adjustments. Speed is excluded from Base SCU and belongs in the live delivery layer.
Frontier-relative SCU
F_SCU_i,t = QualityAdjustedCapability_i,t / FrontierBasket_t
The current top 3 eligible frontier models average to 1.00 F-SCU.
Absolute SCU
Absolute_SCU_i,t = QualityAdjustedCapability_i,t / FrontierBasket_BaseDate
The base date anchor is May 20, 2026. Future frontier baskets can rise above 1 SCU instead of resetting.
Workload Blended Cost
BlendedCost_m = Sum_j workload_weight_j x (input_share_j x input_price_m + output_share_j x output_price_m)
The default standardized workload assumption uses multiple task profiles instead of treating all workloads as a fixed 75/25 token split.
Token-to-SCU Rate
SCU_per_mToken_i,t = Absolute_SCU_i,t x EffectiveTokenEfficiency_i,t
One mToken means one million standard workload tokens. Token efficiency is estimated until direct sampling is available.
Live SCU Spot
Live_SCU_per_1M_tokens_m(t) = Base_SCU_per_1M_tokens_m x DeliveryFactor_m(t) DeliveryFactor_m(t) = ReliabilityFactor x LatencyFactor x AvailabilityFactor x QueueFactor x RoutingFactor
The live layer adjusts the base rate for delivery conditions that can move every 1 to 5 seconds.
Estimated USD per SCU
USD_per_SCU_i,t = BlendedCost_per_mToken_i,t / SCU_per_mToken_i,t
This is a synthetic quote derived from public pricing and estimated capability scores, not a traded market price.
Live USD per SCU
Live_USD_per_SCU_m(t) = Base_USD_per_SCU_m x LatencyPenalty x RetryPenalty x QueuePenalty x CapacityPenalty x RoutingPenalty
Live spot pricing applies real-time delivery penalties to the base capability cost.
Basket USD per SCU
Basket_USD_per_SCU(t) = Sum_m basket_weight_m x Live_USD_per_SCU_m(t)
The primary live quote uses a qualified basket. Best, median, and premium quotes are secondary diagnostics.
SCU Purchasing Power
SCU_per_USD_i,t = SCU_per_mToken_i,t / BlendedCost_per_mToken_i,t
This answers how much estimated capability one dollar can buy.
Capability Oracle
Click each capability block to inspect how it contributes to the Standard Capability Unit.
Limitations
The first SCU release is a research prototype based on estimated and normalized data. Benchmark results may be affected by data contamination, leaderboard overfitting, task mismatch, prompt sensitivity, and provider-specific execution differences.
Before any SCU rate is treated as live, the system should incorporate live benchmark feeds, real-task validation, historical time-series calibration, and transparent methodology notes.