Direct API pricing comparison
Claude Haiku 4.5 vs DeepSeek V4 Pro
Compare current direct API token prices and estimated monthly workload costs for Claude Haiku 4.5 and DeepSeek V4 Pro.
2direct models
Pricing only.DeepSeek V4 Pro is cheaper in 3 of 4 comparable standard workload presets below. This is a pricing comparison only; it does not rank model quality.
| Model | Input | Cached | Output | Context | Max output | Source |
|---|---|---|---|---|---|---|
Claude Haiku 4.5Anthropic · Direct API | $1 | $0.1 | $5 | 200K | 64K | Direct ↗ |
DeepSeek V4 ProDeepSeek · Direct API | $1.32 | $0.044 | $3.96 | 1M | 384K | Direct ↗ |
Same workload, different API bill
These four presets use 30 active days, standard direct rates, no retries, no reasoning-token add-on, no cache writes, and the cached-input percentage shown in each row.
| Standard workload | Claude Haiku 4.5 | DeepSeek V4 Pro | Lower monthly estimate |
|---|---|---|---|
| Chat10,000/day · 2,000 in · 500 out · 25% cached | $1,215 | $1,194.6 | DeepSeek V4 Pro by $20.4 / mo |
| Coding2,000/day · 12,000 in · 1,800 out · 55% cached | $903.6 | $872.78 | DeepSeek V4 Pro by $30.82 / mo |
| Agent1,200/day · 32,000 in · 2,500 out · 65% cached | $928.08 | $921.57 | DeepSeek V4 Pro by $6.51 / mo |
| Documents400/day · 180,000 in · 2,000 out · 15% cached | $1,988.4 | $2,532.82 | Claude Haiku 4.5 by $544.42 / mo |
The main calculator includes batch mode, long-context rules, retries, reasoning tokens, cache writes, and other model-specific pricing logic that a static comparison cannot fully capture.