Compare AI models

Specifications
Parameters
—2.8T
Context window
1M1M
API pricing
Input price
$0.10$3.00
Output price
$0.50$15.00
Cached input price
$0.01$0.30
Cheapest input
$0.10Amazon Bedrock$0.153Wafer
Cheapest output
$0.50Amazon Bedrock$9.50Chutes
$45.00Anthropic list price$535Wafer · mxfp4
Benchmarks
BullshitBench v2
57%74%
Terminal-Bench 4.0
39.2%23.03%
Humanity's Last Exam · no tools
45.9%43.5%
Humanity's Last Exam · with tools
57.4%56%
ARC-AGI-2
61.25%60.4%
threejseval
14241517
Overview
CompanyAnthropicMoonshot AI
Release dateOct 7 2026Jul 16 2026
AccessClosedOpen Weight
Model detailsView modelView model

Frequently asked questions

Claude Haiku 5.5 leads Kimi K3 on 4 of the 6 benchmarks they both report. Claude Haiku 5.5 is cheaper on both input and output: $0.10 vs $3.00 per million input tokens, and $0.50 vs $15.00 per million output tokens. Figures are base-tier rates. Kimi K3 shipped 83 days before Claude Haiku 5.5, so benchmark comparisons should account for the intervening progress.

Context windows are 1M (Claude Haiku 5.5) vs 1M (Kimi K3). Claude Haiku 5.5 is closed, while Kimi K3 is open weight.

On BullshitBench v2, Kimi K3 leads at 74% vs Claude Haiku 5.5 at 57%. On Terminal-Bench 4.0, Claude Haiku 5.5 leads at 39.2% vs Kimi K3 at 23.03%. On Humanity's Last Exam · no tools, Claude Haiku 5.5 leads at 45.9% vs Kimi K3 at 43.5%. On Humanity's Last Exam · with tools, Claude Haiku 5.5 leads at 57.4% vs Kimi K3 at 56%. On ARC-AGI-2, Claude Haiku 5.5 leads at 61.25% vs Kimi K3 at 60.4%. On threejseval, Kimi K3 leads at 1517 vs Claude Haiku 5.5 at 1424.