Pairwise benchmark snapshot · Aug 17, 2026
In this head-to-head showdown, MoonshotAI: Kimi K2.5 delivers higher overall intelligence and human-preferred responses (Elo 1,515 vs 1,224), while MoonshotAI: Kimi K2.5 leads in SWE-bench software engineering benchmarks (71.3%), while MoonshotAI: Kimi K2.5 is more budget-friendly at $2.85/1M per 1M output tokens, while Claude 3.5 Haiku provides faster response streaming (140 tok/s). Review the complete breakdown below to determine which model best fits your performance and budget requirements.
9 benchmarks evaluated across capability, speed, and cost
Previous cheap Claude. Haiku 4.5 is the current $1/$5 SKU.
Auto-discovered from OpenRouter (moonshotai/kimi-k2.5). Preview until a second source matches.
Simulate accuracy gains vs added response time & token cost
Reasoning models “think before answering” by generating internal reasoning tokens. Higher effort improves math, coding, and logic accuracy, but increases response delay and token costs.
Δ 291 Arena Elo pts
140 tok/s
$2.8499999999999996 / 1M output
Relative percentile scores computed across all active models in the benchmark catalog.
Comparing top standard models across 6 key capability dimensions. Tap any spoke or dot to inspect.
Category wins across reasoning intelligence, generation speed, and token cost.
Preference Elo, Coding proficiency, SWE-bench & LiveBench accuracy
Generation throughput and time to first token responsiveness
Cost per million tokens and max context window length
| Benchmark | Claude 3.5 Haiku | MoonshotAI: Kimi K2.5 | Advantage Delta |
|---|---|---|---|
| Preference Elo | +291 | ||
| Coding Elo | — | — | |
| LiveBench | +22.4% | ||
| SWE-bench | +35.1% | ||
| GPQA Diamond | +33.5% | ||
| Time to first token | +160 ms | ||
| Output speed | 140 tok/s | 72 tok/s | +68 tok/s |
| Input price | $0.8/1M | $0.57/1M | +$0.23/1M |
| Output price | $2.85/1M | +$1.15/1M | |
| Context window | +62k |
Simulate monthly production API costs in USD (US Dollar).
MoonshotAI: Kimi K2.5 is estimated to save $25.30/month ($304/year).
Higher SWE-bench (71.3%).
Lower output list price ($2.85/1M).
Lower TTFT (110 ms).
Larger window (262k).
Both accept images. Defaulting to the higher-Elo side (MoonshotAI: Kimi K2.5).
| Target Workload | Recommended Pick | Evaluation Rationale |
|---|---|---|
| Repo / coding agents | MoonshotAI: Kimi K2.5 | Higher SWE-bench (71.3%). |
| High-volume chat | MoonshotAI: Kimi K2.5 | Lower output list price ($2.85/1M). |
| Voice / low-latency UI | Claude 3.5 Haiku | Lower TTFT (110 ms). |
| Long-document RAG | MoonshotAI: Kimi K2.5 | Larger window (262k). |
| Screenshots / vision | MoonshotAI: Kimi K2.5 | Both accept images. Defaulting to the higher-Elo side (MoonshotAI: Kimi K2.5). |
No comments posted on this matchup yet. Be the first to share an evaluation note!
Dynamic OpenGraph banner generated at /compare/claude-3-5-haiku-vs-kimi-k2-5/opengraph-image
Verified Head-to-Head Telemetry
