Claude Haiku 5.5 vs Kimi K3
Claude Haiku 5.5 leads Kimi K3 by 2.6 points on the Shogo Agent Index, but their ranges overlap, so treat it as close. Claude Haiku 5.5 wins 12 and Kimi K3 wins 8 of 20 shared benchmarks. Claude Haiku 5.5 is cheaper for agent workloads.
Add more models to this comparison →Pillars
| Pillar | Claude Haiku 5.5 | Kimi K3 |
|---|---|---|
| Agentic coding | 67 | 63 |
| Long-horizon autonomy | 62 | 65 |
| Knowledge-work agents | 46 | 42 |
| Computer use | 47 | 43 |
| Security & context learning | 55 | 53 |
Shared benchmarks
| Benchmark | Claude Haiku 5.5 | Kimi K3 |
|---|---|---|
| APEX-Agentsindex | 56.1% | 47.8% |
| Vending-Bench 2index | $3,602 | $5,165 |
| FrontierSWEindex | 43.8% | 25.9% |
| ARC-AGI | 89.0% | 94.5% |
| ARC-AGI-2 | 61.3% | 60.4% |
| Chess Puzzles | 55.0% | 37.0% |
| CritPt | 22.6% | 23.4% |
| DTBench | 95.0% | 91.2% |
| FrontierCode | 46.4% | 44.2% |
| FrontierMath Tier 4 | 46.3% | 39.0% |
| FrontierMath Tiers 1-3 | 75.1% | 72.2% |
| GPQA Diamond | 89.6% | 93.1% |
| LMCA | 51.1% | 52.7% |
| MirrorCode | 78.9% | 54.4% |
| OTIS Mock AIME | 98.9% | 97.2% |
| ProofBench | 89.0% | 87.0% |
| SciCode | 55.0% | 59.5% |
| SimpleBench | 59.2% | 60.7% |
| SimpleQA Verified | 23.8% | 50.6% |
| WeirdML v3 | 11.7% | 7.5% |
Specs and price
| Claude Haiku 5.5 | Kimi K3 | |
|---|---|---|
| Agent Index | 57.5 (#18) | 54.9 (#26) |
| Agent $/M | $0.050 | $1.39 |
| Input / output $/M | $0.10 / $0.50 | $3.00 / $15.00 |
| Context | 1M | 1M |
| Released | Oct 7, 2026 | Jul 16, 2026 |
| Weights | Closed | Open |
| API hosts | 5 | 29 |
Don’t pick one. Route.
Shogo can run Claude Haiku 5.5 and Kimi K3 side by side and send each step of an agent to the one that’s cheaper for it.
Try both in Shogo