Pairwise benchmark snapshot · Aug 16, 2026
In this head-to-head showdown, Claude Sonnet 5 delivers higher overall intelligence and human-preferred responses (Elo 1,556 vs 1,438), while Claude Sonnet 5 leads in SWE-bench software engineering benchmarks (73.8%), while Claude Haiku 4.5 is more budget-friendly at $5/1M per 1M output tokens. Review the complete breakdown below to determine which model best fits your performance and budget requirements.
10 direct benchmark disciplines evaluated across capability, speed, and cost
Anthropic cheap/fast Claude SKU. Official Claude Platform list price $1/$5 per 1M tokens (Anthropic Haiku page).
Anthropic workhorse. Official $2/$10 per 1M tokens made permanent on Aug 10 2026.
Currently comparing Claude Haiku 4.5 vs Claude Sonnet 5. Alternative family tiers available.
Δ 118 Arena Elo pts
175 tok/s
$5 / 1M output
Relative percentile scores computed across all active models in the benchmark catalog.
Comparing top Western standard models with China's leading frontier rival across 6 skill 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 Haiku 4.5 | Claude Sonnet 5 | Advantage Delta |
|---|---|---|---|
| Preference Elo | +118 | ||
| Coding Elo | +116 | ||
| LiveBench | +8% | ||
| SWE-bench | +15.6% | ||
| GPQA Diamond | +8.8% | ||
| Time to first token | +80 ms | ||
| Output speed | 175 tok/s | 118 tok/s | +57 tok/s |
| Input price | +$1/1M | ||
| Output price | +$5/1M | ||
| Context window | +800k |
Simulate monthly production API costs in USD (US Dollar).
Claude Haiku 4.5 is estimated to save $110.00/month ($1,320/year).
Higher SWE-bench (73.8%).
Lower output list price ($5/1M).
Lower TTFT (95 ms).
Larger window (1M).
Both accept images. Defaulting to the higher-Elo side (Claude Sonnet 5).
| Target Workload | Recommended Pick | Evaluation Rationale |
|---|---|---|
| Repo / coding agents | Claude Sonnet 5 | Higher SWE-bench (73.8%). |
| High-volume chat | Claude Haiku 4.5 | Lower output list price ($5/1M). |
| Voice / low-latency UI | Claude Haiku 4.5 | Lower TTFT (95 ms). |
| Long-document RAG | Claude Sonnet 5 | Larger window (1M). |
| Screenshots / vision | Claude Sonnet 5 | Both accept images. Defaulting to the higher-Elo side (Claude Sonnet 5). |
$1/$5 hosted Claude vs a cheap long-context open sibling. The alternative-to-Sonnet pair that is honest.
Usually Sonnet 5 vs Opus 5. Coding delta versus output $/1M. The workhorse question.
Aug 10 2026 made the $2/$10 list permanent. Sonnet is the default Claude unless the repo is actually on fire.
Huge window, cheap hosted lists. The Haiku alternative that is actually open-weight.
No comments posted on this matchup yet. Be the first to share an evaluation note!
Dynamic OpenGraph banner generated at /compare/claude-haiku-4-5-vs-claude-sonnet-5/opengraph-image
Verified Head-to-Head Telemetry
launch · Oct 22, 2025
Claude Haiku 4.5 at $1/$5: the cheap Claude people actually ship
compare · Aug 16, 2026
Haiku 4.5 vs Llama 4 Scout: cheap Claude versus cheap open Llama
compare · Aug 10, 2026
Sonnet vs Opus in 2026: same family, different invoice
price · Aug 10, 2026
Claude Sonnet 5 at $2/$10: the 2026 Anthropic workhorse