Launch · Tech Times ·

Kimi K3 adds Standard and High reasoning tiers, easing its max-only cost problem

Two days after launching Kimi K3, the largest open-weight model released to date, Moonshot AI has begun rolling out Standard and High reasoning-effort tiers alongside the existing Max-only mode. The update targets developers who saw per-task costs spike since K3 launched with only maximum-effort reasoning available.

Based on reporting by Tech Times — analysis by dalili

Kimi K3 launched July 16 as a 2.8-trillion-parameter Mixture-of-Experts model — roughly 75% larger than DeepSeek's V4 Pro — with a 1-million-token context window and native multimodal input, landing at the frontier tier on several independent benchmarks while trailing Claude Fable 5 and GPT-5.6 Sol on broad measures. At launch, the model exposed only a single reasoning_effort setting: max, always on, with reasoning tokens billed as output at $15 per million tokens.

That single-tier design proved expensive fast: independent testing found a simple SVG-generation prompt consuming over 13,000 reasoning tokens to produce roughly 3,400 tokens of actual output, costing around $0.25 for a trivial request. Moonshot's updated Kimi Code documentation, published July 18, now maps all three effort levels — Standard, High, and Max — with the rollout reaching the Kimi Code environment first; the main API quickstart still lists only max as supported, suggesting a staged release.

The addition gives developers their first real lever on the speed-versus-compute tradeoff for a model Moonshot has positioned as evidence that open-weight models are closing the gap with the closed Western frontier — full open weights for K3 are dated for July 27.

Key takeaways

  • Kimi K3 launched July 16 with only max-effort reasoning, driving heavy per-task costs
  • July 18 update adds Standard and High tiers, first rolling out in the Kimi Code environment
  • Full open weights for the 2.8T-parameter model are dated for July 27

Why it matters

A frontier-tier open-weight model that only ran at maximum cost was a barrier to real adoption — tiered reasoning effort is the difference between an impressive benchmark result and a model teams can actually afford to run in production.