Compare verified published API prices for Kimi K3 and DeepSeek V4 Pro. This page compares price, not output quality.
Verified Oct 9, 2026, Kimi K3 lists input at $3.00 and output at $15.00 per 1M tokens, while DeepSeek V4 Pro lists $1.32 input and $3.96 output. DeepSeek V4 Pro is 56% cheaper on input. DeepSeek V4 Pro is 73.6% cheaper on output. This is a comparison of published prices only, not output quality.
| Model | Input $/1M | Cached input $/1M | Output $/1M | Source | Last verified |
|---|---|---|---|---|---|
| Kimi K3 | $3.00 | $0.30 | $15.00 | Official source | Sep 29, 2026 |
| DeepSeek V4 Pro | $1.32 | $0.044 | $3.96 | Official source | Oct 9, 2026 |
Prices from official provider pages, verified on the date shown.
Published rates
| Field | Kimi K3 | DeepSeek V4 Pro |
|---|---|---|
| Provider | Moonshot AI | DeepSeek |
| Input / 1M | $3.00 | $1.32 |
| Cached input / 1M | $0.30 | $0.044 |
| Output / 1M | $15.00 | $3.96 |
| Context window | 1,048,576 | 1M |
| Release date | Jul 16, 2026 | Apr 24, 2026 |
| Official source | Moonshot AI pricing ↗ | DeepSeek pricing ↗ |
Standard token prices
Costs below allocate 1M total tokens across each input:output ratio. They use standard prices only, without cache or Batch discounts.
| Input:output ratio | Kimi K3 | DeepSeek V4 Pro | Cheaper published price |
|---|---|---|---|
| 10:1 (RAG / long context) | $4.09 | $1.56 | DeepSeek V4 Pro by 61.9% |
| 3:1 (chat) | $6.00 | $1.98 | DeepSeek V4 Pro by 67% |
| 1:1 | $9.00 | $2.64 | DeepSeek V4 Pro by 70.7% |
| 1:3 (long generation) | $12.00 | $3.30 | DeepSeek V4 Pro by 72.5% |
Computed examples
| Workload | Assumed volume | Kimi K3 | DeepSeek V4 Pro | Difference |
|---|---|---|---|---|
| Side project | 10,000 requests/month | $69.00 | $22.44 | DeepSeek V4 Pro by $46.56 (67.5%) |
| Customer support chatbot | 10,000 conversations/month | $526.95 | $203.35 | DeepSeek V4 Pro by $323.60 (61.4%) |
| Agent task | 1,000 tasks/month | $84.00 | $28.51 | DeepSeek V4 Pro by $55.49 (66.1%) |
Provider conditions
Official pricing notes: Automatic prefix caching uses the cache-hit rate for eligible repeated prefixes of more than 256 tokens. K3 reasoning is always enabled; reasoning tokens are billed as output tokens. No current K3 Batch price is published in the canonical contract.
Rates are selected by request time in UTC.
Official pricing notes: Higher-capability DeepSeek tier with 1M context and 384K max output. DeepSeek uses time-based pricing. This estimate uses the peak rate; off-peak rates are 50% lower.
Operational costs
Retries, failed calls, fallback requests, and longer outputs can raise the cost of completing a task even when the published token price is lower. Model the effect with the LLM Retry & Fallback Cost Calculator, read the cost-per-successful-task guide, or estimate a migration with the Model Switch Cost Calculator.
Primary evidence and next steps
Common questions
DeepSeek V4 Pro is 56% cheaper on input. DeepSeek V4 Pro is 73.6% cheaper on output. This compares published list prices, not output quality.
At 800 input and 300 output tokens per request, Kimi K3 costs about $69.00 and DeepSeek V4 Pro costs about $22.44 at standard list prices.
At a 10:1 input-to-output token ratio, Kimi K3 costs $4.09 and DeepSeek V4 Pro costs $1.56 per 1M total tokens. DeepSeek V4 Pro by 61.9%.
The prices come from Moonshot AI and DeepSeek official pricing sources, verified on Sep 29, 2026 and Oct 9, 2026, respectively. See the AI API pricing dataset.