Kimi K3
Last verified August 3, 2026 · Moonshot pricing ↗$3.00/1M input · $15.00/1M output · 1.0M context · Moonshot
Count Tokens
—
Tokens
—
Input Cost
—
Output Cost
Estimate Monthly Cost
Monthly Cost Estimator
Quick:
<$0.0001/mo
Pick a preset above or enter custom usage
Alternatives to Kimi K3
Pricing Details
Moonshot2.8T param MoE (16 of 896 experts active), native vision, always reasons. Cached input: $0.30/1M. Full 1M context billed flat with no length tiering. Max output defaults to 131,072 and can be raised to the full 1,048,576. Rate card is 3.16x input and 3.75x output over K2.6. AA Intelligence Index 57, rank #4 (Jul 17, 2026), at $0.94/task vs Opus 4.8's $1.80; #1 on Arena Frontend Code (1679). Very verbose: 130M tokens across AA's suite vs 63M average, though ~21% fewer output tokens than K2.6. AA measured hallucination rising to 51% from K2.6's 39%. Open weights released 2026-07-27 on HF (moonshotai/Kimi-K3), ungated, modified-MIT (separate agreement required for MaaS operators above $20M/12mo revenue). Checkpoint is 1,560.9 GB in MXFP4 across 96 shards; vLLM requires 8x B300 / 16x B200 minimum (8x MI355X also fits), so no host undercuts the first-party rate on synchronous requests: Fireworks, Together, Baseten, SiliconFlow, Novita and Atlas Cloud all match $3.00/$15.00/$0.30. Nebius matches $3/$15 with no cache discount ($4.20 blended). Only sub-list price anywhere is Fireworks batch at $1.50/$7.50 (platform-wide 50% async modifier). Note K3's license adds a MaaS clause K2's lacked. API ID: kimi-k3.
Input / 1M tokens
$3
Output / 1M tokens
$15
Context Window
1.0M
Max Output
131K
Price History
No price change recorded since 2026-08-03, when this site started tracking this model; it launched on 2026-07-16.
Frequently Asked Questions
Common questions about Kimi K3 pricing and usage
Read More About Kimi K3
Anthropic counted 189.9 million distillation exchanges across five Chinese labs and printed no token count and no dollar figure. At the Opus card the campaign is worth between $5.7M and $40M, and every Kimi request Moonshot relayed to Opus would have lost 40 cents on the dollar if anyone had paid for it.
September 14, 2026 · 13 min read
Sakana priced Fugu Max at $2 and $6, which is Qwen3.8 Max's card to the cent, and says the output line is 40% to 60% under Sonnet 5, Terra and Kimi K3. The tokens billed at that rate include every internal call the orchestrator makes, and Sakana publishes neither the count nor the models it calls.
September 12, 2026 · 14 min read
GLM-5.3 bills the same $1.40 and $4.40 that GLM-5.2 and GLM-5.1 billed, and finishing a task still costs 55% more than it did. The rate card is the wrong number to be watching.
August 22, 2026 · 11 min read