THE AI COMPUTE PRICING ENGINE
One API. All models. A benchmark-pegged Algorithmic Pricing Engine that recalculates rates daily based on verified performance tests. Route requests to Claude, GPT, and Gemini through a single endpoint — priced algorithmically by benchmark scores.
HOW IT WORKS
Add credits, pick a model, and compute. Route requests to any supported LLM at algorithmically-determined prices.
ADD CREDITS
Fixed peg. Zero volatility.
Add credits via Stripe. 1 credit = $0.01 USD — always. This is your base currency on the De.Ciphyr platform.
CHOOSE YOUR MODEL
Benchmark-verified pricing.
Pick the best model for your task. Pricing adjusts daily based on automated benchmark performance scores — so you always know you're getting fair value.
CONNECT & COMPUTE
One endpoint. Infinite models.
Use a single unified API endpoint with dynamic routing, or create specific endpoints for individual LLM tokens. Drop-in compatible with OpenAI SDKs.
No credit card required to explore pricing
ANTI-GAMING
EVALUATION ENGINE
Goodhart's Law states: "When a measure becomes a target, it ceases to be a good measure." De.Ciphyr defends against benchmark gaming with multi-axis evaluation, hidden test sets, and adversarial contamination detection.
Multi-Axis Composite Scoring
30% Hidden Test Sets
Proprietary, rotating evaluation sets never published. Refreshed monthly to prevent data leakage and training contamination.
Adversarial Prompts
20% of evaluation uses paraphrased variants of known benchmarks to detect memorization and overfitting.
Contamination Detection
Statistical analysis of response patterns flags verbatim reproduction of benchmark answers, triggering review and potential delisting.
ADVANCED COMPUTE
OPTIMIZATION
LIMIT ORDERS
Automate your infrastructure costs. Set a limit order to automatically buy 1M Claude tokens if the price drops below 50 tokens/credit. Lock in favorable rates automatically.
VERIFIABLE BENCHMARKS
Multi-axis evaluation across Code, Reasoning, Tool Use, and Instruction Following — with 30% hidden test sets rotated monthly. Price reflects verified performance, not marketing claims.
CIRCUIT BREAKERS
Automated rate protection. If any model's pricing swings more than 25% in a single recalibration, routing halts for that model pending review. Enterprise-grade stability.
COMPUTE RESELLING
Enterprise-grade flexibility. Soon, enterprises will be able to lock in compute rates for future usage, and independent providers will be able to list their models on the platform.
MODEL PRICING
Rate = tokens per 1 credit · Benchmarked daily · Backed by USD deposits via Stripe
AI MODELS
Benchmark-driven pricing across providers. 1 Credit = $0.01 USD. Rates update daily based on automated evaluations across code generation, reasoning, tool use, and instruction following.
Sorted by quality score
| # | Model | Provider | Benchmark | Tokens/Credit | Input $/M | Output $/M | Tier |
|---|---|---|---|---|---|---|---|
| #1 | Claude Opus 4.7 claude-opus-4.7 | anthropic | 93 | 59 | $5.00 | $25.00 | Premium |
| #2 | o3 o3 | openai | 92 | 161 | $2.00 | $8.00 | Premium |
| #3 | GPT-5.5 gpt-5.5 | openai | 91 | 45 | $5.00 | $30.00 | Premium |
| #4 | Gemini 3.1 Pro gemini-3.1-pro | 90 | 109 | $2.00 | $12.00 | Premium | |
| #5 | Claude Sonnet 4.6 claude-sonnet-4.6 | anthropic | 88 | 91 | $3.00 | $15.00 | Premium |
| #6 | Gemini 2.5 Pro gemini-2.5-pro | 87 | 139 | $1.25 | $10.00 | Premium | |
| #7 | DeepSeek V4 Pro deepseek-v4-pro | deepseek | 86 | 333 | $1.74 | $3.48 | Premium |
| #8 | GPT-5.4 gpt-5.4 | openai | 85 | 91 | $2.50 | $15.00 | Premium |
| #9 | Grok 4.3 grok-4.3 | xai | 84 | 500 | $1.25 | $2.50 | Premium |
| #10 | o4-mini o4-mini | openai | 83 | 294 | $1.10 | $4.40 | Premium |
Benchmark scores are recalculated daily at 00:00 UTC across code generation, reasoning, tool use, and instruction following tasks.
BUILT FOR
INSTITUTIONAL CREDIBILITY
De.Ciphyr operates with the transparency and risk controls expected by enterprise buyers. Algorithmic pricing, circuit breakers, and formal risk disclosures — not just promises.
TRANSPARENT PRICING
Every pricing rate is derived from verified benchmark scores using a published algorithm. No hidden markups. No opaque pricing tiers. Full auditability.
CIRCUIT BREAKERS
If any model's pricing rate swings more than 25% in a single recalibration cycle, routing is automatically halted for that model pending manual review.
BENCHMARK INTEGRITY
Multi-dimensional scoring across Code, Reasoning, Tool Use, and Instruction Following — with 30% hidden test sets rotated monthly to prevent gaming.
PROVIDER FALLBACK
Automatic failover to alternative providers if the primary model endpoint is unavailable. Multi-provider routing ensures your requests always complete.
Pricing Rate Variability
Model pricing rates may change due to provider pricing changes, benchmark methodology updates, or new model releases.
Benchmark Methodology Risk
While De.Ciphyr employs multi-axis hidden-set evaluation, no benchmark methodology is immune to all forms of gaming. The evaluation approach continuously evolves.
Provider Dependency
The platform depends on upstream provider API availability. Provider outages, rate limiting, or silent model swaps may affect service quality.
Credit Balance Risk
Credits represent prepaid compute access. Unused credits do not accrue interest or appreciate in value. Credits are non-refundable except as provided in the Terms of Service.
Credits represent prepaid compute access, not securities, currencies, or investment instruments. This is not investment advice. See our documentation for full risk disclosures and operational details.