LLM API Cost Calculator

What does your AI agent actually cost?

Estimate the per-token API bill for any frontier model at your real usage — using live OpenRouter pricing — and see it against Standard Compute's flat monthly price. Agent workloads are where per-token gets brutal.

Claude Opus 4.8 · per-token
$4,500/mo
90,000 requests/mo
vs
Standard Compute · flat
$39/mo
unlimited · from $9/mo
That's $4,461/mo saved — about 115× cheaper at this usage.

Every model at this usage

Monthly per-token cost, cheapest first. Standard Compute is flat at $9–$399 regardless of volume.

ModelPer-token / movs Standard ($39)
Qwen3 235B A22B · Alibaba (Qwen)$501.3× more
DeepSeek V4 Flash · DeepSeek$571.5× more
Llama 4 Maverick · Meta$1223.1× more
DeepSeek V3.2 · DeepSeek$1343.4× more
MiniMax M2.1 · MiniMax$2165.5× more
Codestral · Mistral$2165.5× more
MiniMax M3 · MiniMax$2436.2× more
Gemini 3.1 Flash Lite · Google$2476.3× more
Qwen3.7 Plus · Alibaba (Qwen)$2596.6× more
DeepSeek V4 Pro · DeepSeek$2747.0× more
GLM 4.7 · Z.AI$3388.7× more
GLM 4.6 · Z.AI$3509.0× more
Mistral Large 3 · Mistral$3609.2× more
Devstral 2 · Mistral$3609.2× more
DeepSeek R1 (0528) · DeepSeek$41811× more
Kimi K2.7 Code · Moonshot AI$55214× more
Kimi K2.6 · Moonshot AI$60415× more
GLM 5.2 · Z.AI$71818× more
GPT-5.4 Mini · OpenAI$74319× more
Grok 4.3 · xAI$78820× more
Grok 4.20 · xAI$78820× more
Claude Haiku 4.5 · Anthropic$90023× more
Qwen3.7 Max · Alibaba (Qwen)$90023× more
Gemini 3.5 Flash · Google$1,48538× more
Gemini 3.1 Pro · Google$1,98051× more
GPT-5.3-Codex · OpenAI$2,04853× more
GPT-5.4 · OpenAI$2,47563× more
Claude Sonnet 4.6 · Anthropic$2,70069× more
Claude Opus 4.8 · Anthropic$4,500115× more
GPT-5.5 · OpenAI$4,950127× more

Estimates use public OpenRouter per-token pricing (input + output) at the usage above. Real bills vary with caching, retries, and tool calls — which agents do constantly, so per-token tends to run higher than estimated. Standard Compute pricing is flat and unlimited within fair use.

Stop watching the meter

Point any OpenAI-compatible agent at Standard Compute and the bill stops scaling with usage. One flat price, unlimited compute, auto-routed frontier models.

Get a free API key →See how it connects →