Skip to content

Kimi K3 vs Claude Opus 5.5

Kimi K3 is from Moonshot AI, released July 16, 2026. Claude Opus 5.5 is from Anthropic, released September 22, 2026.

  • LMArena Elo1488 vs 1509. Claude Opus 5.5 is 1.4% ahead on LMArena Elo.
  • Input per 1M$3 vs $4. Kimi K3 costs 25% less per million input tokens.
  • Output per 1M$15 vs $20. Kimi K3 costs 25% less per million output tokens.
  • Context window1M vs 1M. Both hold 1M tokens of context.

Prices are the vendors’ published API rates; Elo is LMArena’s text board. Each figure is dated on its model page.

Interactive Comparison

Loading chart...
$3 / $15 per million tokens
$4 / $20 per million tokens
Identity
DeveloperMoonshot AIAnthropic
ReleasedJul 2026Sep 2026
StatusGAGA
LicenceOpen weights (Kimi K3 License)Proprietary
Self-hostableYesNo
Cost
Blended $/1M tokensinput × 0.75 + output × 0.25$6.000 / 1M tokens$8.000 / 1M tokens
Input price$3.000 / 1M tokens$4.000 / 1M tokens
Output price$15.000 / 1M tokens$20.000 / 1M tokens
Cached input$0.30 / 1M tokens (automatic context-cache hit; $3.00 on a cache miss)$0.20 per 1M tokens
Batch discountNot offered50% ($2 / $10 per 1M tokens)
Free tierNot offered—
Capacity
Context window1M tokens1M tokens
Max output128k tokens (default; max_completion_tokens configurable up to 1M)128k tokens
Long-context surchargeNone (flat rate across the 1M window)—
Capability
Vision inYesYes
Audio inNoNo
Function callingYesYes
Structured outputYesYes
Extended reasoningYesYes
Web searchYes (tool currently being updated by Moonshot; not recommended for near-term use)Yes
Code executionNot offeredYes
Access
APIYesYes
Chat appYesYes
Cloud marketplacesAzure AI Foundry (via Fireworks AI), OpenRouterAmazon Bedrock, Google Cloud, Microsoft Foundry
Fine-tuningNot offeredNo
Measured quality
LMArena Elo1488 (checked Nov 2026)1509 (checked Nov 2026)
LMArena rankRank 19Rank 3
GPQA Diamond9391 · max effort
Elo per dollarLMArena Elo ÷ blended $/1M tokens248189

Add a third model to this comparison