Kimi API pricing by model and provider

Route-specific buying guide

How to budget Kimi K3 on OpenRouter

Route ID: moonshotai/kimi-k3

The accepted rows describe OpenRouter's Kimi K3 route. They are not Moonshot's direct developer prices and do not establish a universal cheapest Kimi provider.

OpenRouter reported a 1,048,576-token context window when checked. Context capacity is a ceiling, not a budget target; retrieval and conversation growth can turn input tokens into the dominant cost.

Cost formula

Estimate one request as uncached input tokens × the input rate, cached input tokens × the cached-input rate, plus output tokens × the output rate. Divide every token count by one million before applying APIDir's displayed rates.

Costs that move the bill

• Long documents, retrieval payloads, conversation history, and agent memory
• Completion length and repeated reasoning, coding, browsing, or tool-use loops
• Cache coverage, cache invalidation frequency, and any route-specific write charge
• Retries, timeouts, fallback models, rate limits, and production support needs

Checks before purchase

• Keep the moonshotai/kimi-k3 route ID and the checked date beside every cost result.
• Benchmark short, median, and long prompts because a single average hides tail cost.
• Cap output and agent steps deliberately, then measure successful-task cost rather than request cost.
• Treat OpenRouter and Moonshot direct access as different offers until their commercial boundaries match.

Source checked Aug 22, 2026: OpenRouter public model API. Recheck the live route before committing spend.

Estimate this route with the LLM workload calculator →

Compare another model family

• Qwen API pricing
• Grok API pricing
• Claude API pricing

Decision guide

How to compare Kimi API providers

Kimi is a model family, so a useful provider comparison must name the exact release. The reviewed ranking uses Kimi K3 across OpenRouter, Together AI, and Fireworks AI. Each provider's route spelling remains visible even when the underlying release and token meter match.

Equal token rates do not prove equal service. Context support, cache behavior, throughput, quotas, fallback, endpoint compatibility, regional access, data terms, and support remain separate buying criteria. Use the ranking as a price-meter result and benchmark the service characteristics yourself.

What changes Kimi API cost

• Exact Kimi release, input-output ratio, context, and cache use
• Router or aggregator tier, throughput, rate limits, and any dedicated option
• Tool calls, retries, data retention, regional access, and support

Integration checks before production

• Pin Kimi K3's full provider route ID and reject silent fallback
• Set context, output, timeout, and retry budgets at the caller boundary
• Log token counts, route, terminal status, retry count, and request ID

Workloads this comparison can inform

Long-context assistants

Verify context support and cache meters before extending a simple token estimate.

Batch language processing

Measure real output ratios and concurrency instead of ranking input price alone.

Multi-provider routing

Test fallback identity and error semantics even when the headline rates tie.

How to read the price result

Three reviewed Kimi K3 routes currently publish the same standard input and output rates. APIDir reports the tie for those meters only; it is not a quality, speed, uptime, or support ranking.

Related model price comparisons

• Qwen API pricing comparisonCompare Qwen API pricing by exact model ID, provider channel, input and output token meter, context, and official source.
• GLM-5.2 pricing comparisonCompare GLM-5.2 API token prices across provider routes with exact model identity, meter, tier, source, and check date.
• Claude API pricing comparisonCompare Claude API routes by exact model, context tier, input and output meter, direct or router channel, source, and effective date.