Dual-stage token savings calculator
Calculate your monthly savings
Optimize input prompt bloat and eliminate expensive output generation tokens. Select your monthly volume and production model to inspect your net ROI.
M tokens
Presets:
Production model
Input payload architecture
-60.5% input tokensOutput response efficiency
-35% output tokensSub-2ms latency
Zero schema syntax drift
1-line baseURL proxy drop-in
Estimated net savings
$13,021/mo
+$156,252 projected annual savings
Current provider bill$30,000/mo
New bill with Curtly$16,950/mo
Curtly planGrowth ($29/mo)
Input prompt savings (60.5%)-$6,050/mo
Output generation savings (35%)-$7,000/mo
Payback return44,900% ROI (pays off in < 24h)
Start optimizing freeInspect benchmark comparisons below↓
Free tier available. No credit card required.
Explore model-specific token calculators
AnthropicClaude BPE
Claude Opus 5
$5.00 in · $25.00 out (5x)Calculate
AnthropicClaude BPE
Claude Fable 5.1
$10.00 in · $50.00 out (5x)Calculate
OpenAIo200k_base
GPT 6 Astra
$10.00 in · $50.00 out (5x)Calculate
DeepSeek128k BPE
DeepSeek R1
$0.55 in · $2.19 out (4x)Calculate
Moonshot AIKimi 160k BPE
Kimi K3
$3.00 in · $15.00 out (5x)Calculate
AnthropicClaude BPE
Claude Mythos Preview
$10.00 in · $50.00 out (5x)Calculate
OpenAIo200k_base
GPT-5.6 Sol
$4.00 in · $20.00 out (5x)Calculate
Zhipu AIGLM 151.5k BPE
GLM-5.3
$1.40 in · $4.40 out (3.1x)Calculate
AlibabaQwen 152k BPE
Qwen 3.8 Max
$2.00 in · $6.00 out (3x)Calculate
TencentHunyuan 128k BPE
Hunyuan Frontier
$0.83 in · $2.50 out (3x)Calculate
Meta128k BPE
Muse Spark 1.3
$1.25 in · $4.25 out (3.4x)Calculate
OpenAIo200k_base
GPT-5.6 Terra
$2.00 in · $12.00 out (6x)Calculate
OpenAIo200k_base
GPT-5.6 Luna
$0.20 in · $1.20 out (6x)Calculate
Architecture comparison
Why engineering teams and CTOs choose Curtly
Python-heavy minifiers require local GPU hosting and risk JSON/code syntax corruption. Curtly is a zero-maintenance drop-in gateway that safely governs your entire LLM request/response lifecycle in under 2ms.
| Capability / Dimension | Curtly (Two-Stage) | LLMLingua |
|---|---|---|
Optimization coverage | Input + output (two-stage) | Input only |
Code & schema safety | Zero corruption guarantee | Frequent syntax breakage |
Integration effort | 1-line baseURL swap | Heavy Python + PyTorch dependency |
Hardware & compute | 0 MB local compute required | Requires local GPU & 7B model |
Latency overhead | Under 15ms proxy routing | 300ms to 800ms overhead |
Total cost reduction | 55% to 70% total saved | 30% to 40% saved |
Benchmark scenario inspector
Compare real token savings and syntax safety on realistic developer prompts.
Agent memory & state
Initial input size: 382 tokens
Curtly Safe Vault™
-57% savedOriginal baseline
382 tokens
Optimized prompt
164 tokens-57%
JSON schema preserved bit-for-bit without quote loss
Bit-for-bit syntax protection verified for code & JSON
Microsoft LLMLingua
-43% savedOriginal baseline
382 tokens
Optimized prompt
218 tokens-43%
Corrupted JSON syntax and missing brackets
High risk of broken quotes, stripped keywords or bracket loss
Ready to cut your company's AI infrastructure costs?
Start with 500 free requests per month. Connect your production app in 2 minutes.
FAQ
Straight answers. No fluff.
Curtly works with any LLM. Optimization operates across two layers: dynamic input compression before network dispatch, and output efficiency directive injection (Curtly Brevity Engine). Whether you route to OpenAI, Anthropic Claude, Groq, OpenRouter, or custom upstream endpoints, Curtly optimizes both the prompt payload and the generated response.