Kimi K3.1 Pro Pricing: Complete API Token Rates and Cost Breakdown

Kimi K3.1 Pro costs $2.00 per million input tokens, $10.00 per million output tokens, and $0.20 per million cached input tokens on the pay-as-you-go Kimi application programming interface (API). The faster variant, Kimi K3.1 Pro Highspeed, costs $4.00 per million input tokens and $20.00 per million output tokens. Standard K3.1 Pro delivers lower token rates than the older Kimi K3 model, which charges $3.00 for input and $15.00 for output.

At Layer3Labs, we build custom agents and enterprise integrations for business teams. Output tokens cost five times the input rate, so verbose output can raise the bill quickly.

Kimi K3.1 Pro API Pricing

Kimi K3.1 Pro API token pricing is 33 percent cheaper than Kimi K3 across prompt inputs and completions. Moonshot AI lists all rates on a pay-as-you-go basis per one million tokens on its official Kimi API pricing page.

The standard K3.1 Pro model charges $2.00 per million input tokens and $10.00 per million output tokens. If a request hits the prompt cache, that input drops to $0.20 per million tokens.

Moonshot reports that Kimi K3.1 Pro Highspeed delivers 2-3x the speed of K3.1 Pro at a higher price point. All models in the K3 series support a full 1,048,576-token context window.

• kimi-k3.1-pro: $2.00 input, $10.00 output, $0.20 cached input per 1M tokens
• kimi-k3.1-pro-highspeed: $4.00 input, $20.00 output, $0.20 cached input per 1M tokens
• kimi-k3 (predecessor): $3.00 input, $15.00 output, $0.30 cached input per 1M tokens
• All K3 models support a context window of 1,048,576 tokens

Billing Mechanics and Context Caching Rules

The Kimi API is pay-as-you-go with no subscription. Stated prices exclude taxes.

Context caching is automatic for requests that repeat identical prefix tokens. The default cache time-to-live (TTL) is 5 minutes, with an option to extend storage to 1 hour. Cache writes are billed separately depending on the selected TTL duration, while subsequent cache hits receive the $0.20 per million token rate.

File extraction and file storage APIs are temporarily free. Kimi API charges are separate from Kimi membership and Kimi Code plans, whose prices this page does not cover.

• Pay-as-you-go account billing with no subscription
• Automatic context caching on matching prompt prefixes
• Selectable cache TTL settings of 5 minutes or 1 hour
• Separate cache write fees based on chosen TTL duration
• File extraction and file storage endpoints are temporarily free
• Reasoning effort parameters accept low, high, and max configurations

Worked Example: A 10M-Input, 2M-Output Job

An example batch of 10 million input tokens and 2 million output tokens costs $40.00 using standard Kimi K3.1 Pro. The input portion totals $20.00 (10 million multiplied by $2.00 per million), and the completion portion totals $20.00 (2 million multiplied by $10.00 per million).

The same 12-million-token job on Kimi K3.1 Pro Highspeed costs $80.00. That is $40.00 for inputs and $40.00 for outputs. By comparison, processing the same batch on legacy Kimi K3 costs $60.00 ($30.00 for inputs and $30.00 for outputs).

Caching cuts the cost of repeated documents. Say 8 million of the 10 million input tokens hit a warm cache. Input drops from $20.00 to $5.60 ($4.00 fresh plus $1.60 cached). The K3.1 Pro job then costs $25.60 before cache write fees. These figures serve as an arithmetic illustration based on published token prices rather than a formal price quote.

• Standard K3.1 Pro (10M input / 2M output): $40.00 total
• Highspeed K3.1 Pro (10M input / 2M output): $80.00 total
• Predecessor K3 (10M input / 2M output): $60.00 total
• Standard K3.1 Pro with 80% cached input: $25.60 total

Benchmark Token Consumption and Operational Overhead

Artificial Analysis puts the cost of running its Intelligence Index at about $2.11 per task. The model scores 55, ranked fourth of 215.

The model used 150 million output tokens to complete the evaluation run, which Artificial Analysis categorizes as very verbose compared to the index median of 87 million tokens. Because completion tokens cost five times as much as prompt tokens, verbose output raises the bill.

Artificial Analysis recorded output speed at 57.5 tokens per second, which ranks 141 out of 186 models and is characterized as notably slow. The Kimi K3.1 Pro overview covers the model itself, and the Kimi K3 pricing guide lists the predecessor's rates.

• Benchmark testing registered $2.11 average cost per evaluation task
• Intelligence Index score of 55 places fourth among 215 models
• Generated 150M completion tokens, categorized as very verbose against the 87M index median
• Throughput measured at 57.5 tokens per second on standard inference

Steps to Control Production Expenses

Enterprises subject to strict data residency rules should evaluate infrastructure terms before sending proprietary documents to Moonshot AI servers. Zero Data Retention (ZDR) is available for API accounts, and Kimi Business provides team management with separation between personal and company workspaces.

Output tokens cost five times the input rate. The API accepts reasoning_effort values of low, high and max, so run a sample job at each setting and compare output-token counts and cost.

Our pricing verdict would change once Moonshot AI releases the full model weights. Moonshot AI stated in its launch post that full weights are scheduled for release by November 4, 2026. If your team can host the model on local hardware, self-hosting would swap per-token charges for your own hardware and operating costs.

Before you commit, map your expected context sizes against the rates on the Kimi API pricing page to set a token budget.