Claude Haiku 4.5
Anthropic's fastest and cheapest — 61.4% on DualEntry, strong for high-volume classification.
Accounting overall
61.4%
Input / Output
$0.25 / $1.25 per MTok
Context
200K
Speed
~200 tok/s
Released
2025-10
Cutoff
2025-07
Eight accounting-task categories borrowed from DualEntry's 101-task benchmark. Measured where published, synthesized from adjacent benchmarks otherwise.
Haiku 4.5 is the right choice for high-volume, latency-sensitive, narrow-scope accounting workloads — invoice classification, transaction categorization, simple extraction. At 61.4% DualEntry overall, it's not going to autonomously run a month-end close, but its structured-task scores (transaction classification in particular) are competitive, and it's five to twenty times cheaper than Opus 4.7 at the same per-request pricing tier.
The 200K context is the meaningful limiter vs the 1M-context alternatives. For high-volume workflows that don't need long context, that's fine.
Citations
- DualEntry benchmark (Haiku 4.5 61.4%)dualentry.com/blog/claude-opus-4-7-accounting-ai-benchmark-results