Source-checked pricing guides
AI API pricing guides by model and provider
Each guide converts one model or provider's published meters into a cost you can budget: token ratios, per-second video, per-image and per-megapixel rates, cache terms, and the overrides that change the bill. Every rate carries its source and check date, and every guide embeds a calculator that runs the arithmetic on your own inputs.
OpenRouter pricing: routes, token meters, and overrides
How OpenRouter prices a request: route selection, per-million token rates, cache reads, batch variants, and the overrides that change the bill mid-prompt.
Read the 9-minute guide →qwen3.8 max price per million tokensQwen API pricing per million tokens
Qwen3.8 Max, Prime, Flash, and batch route prices with a worked monthly estimate, cache math, and the context limits that decide which Qwen route fits.
Read the 8-minute guide →claude sonnet 5 pricing per million tokensClaude API pricing per token, cache, and batch
Claude Sonnet 5 API pricing per million tokens with cache-write math, cache-read savings, batch rates, and a worked agent workload estimate.
Read the 9-minute guide →kimi k3 pricing per million tokensKimi K3 API pricing and output-token economics
Kimi K3 lists cheap input and expensive output. See the per-million rates, the 95% output share in a realistic workload, and the caps that change the bill.
Read the 8-minute guide →glm-5.2 input and output token priceGLM-5.2 API pricing and workload fit
GLM-5.2's listed input price is a fraction of its output price. See the per-million rates, the workload math, and where the route wins or loses.
Read the 7-minute guide →gpt-5.4 api pricing per million tokensGPT API pricing across the 5.4 family
GPT-5.4 family prices per million tokens, the 272,000-token override, batch and cache rates, and a worked estimate for long-document workloads.
Read the 9-minute guide →seedance 2.0 api pricing per secondSeedance 2.0 API pricing by provider and unit
Seedance 2.0 prices differ by provider and by billing unit. Compare per-second, per-token, and per-video rates with worked clip math and retry costs.
Read the 9-minute guide →kie.ai credits to usdKie.ai credit pricing and dollar conversion
Kie.ai bills in credits worth about half a cent each. See the conversion, per-model credit rates, and a worked monthly budget for image and video batches.
Read the 8-minute guide →flux image api pricing per megapixelFLUX image API pricing per image and per megapixel
FLUX API prices depend on whether the provider bills per image or per megapixel. Compare routes, resolution math, and monthly batch cost.
Read the 8-minute guide →gpt image 2 price per imageGPT Image 2 pricing per image and per million tokens
GPT Image 2 costs $0.025-$0.03 per image on Kie.ai and $0.047-$0.053 on Replicate, fal and Together AI. One tier bills per million tokens.
Read the 8-minute guide →How these guides stay honest
APIDir does not claim a universal cheapest provider. A rate is published with its provider, configuration, unit, source URL, and check date, and a rate whose freshness window has expired is treated as unverified rather than current. When a retrieved provider page disagrees with an earlier accepted observation, the guide shows both and says which one the budget should use.
Calculators use the rates you enter, so an old number cannot silently become a recommendation. If a guide's rate changes, the source list is where the correction starts.