Anthropic: Claude Haiku 4.5 vs Microsoft: Phi 4
Microsoft: Phi 4 is cheaper than Anthropic: Claude Haiku 4.5 by about 96% on a blended (75% input / 25% output) basis. Anthropic: Claude Haiku 4.5 charges $1.000000 / $5.000000 per 1M input/output tokens; Microsoft: Phi 4 charges $0.070000 / $0.140000.
| Field | Anthropic: Claude Haiku 4.5 | Microsoft: Phi 4 |
|---|---|---|
| Input $/Mtok | $1.000000 | $0.070000 |
| Output $/Mtok | $5.000000 | $0.140000 |
| Cache read $/Mtok | $0.100000 | – |
| Cache write $/Mtok | $1.250000 | – |
| Input $/Mchar | n/a | n/a |
| Output $/Mchar | n/a | n/a |
| Context window | 200000 | 16384 |
| Tokenizer | claude | – |
Price history
Same workload, monthly cost
Illustrative monthly cost at 10M tokens/month, split per workload preset. Not a substitute for the calculator, which uses your own volume, cache and batch assumptions.
| Workload | Anthropic: Claude Haiku 4.5 | Microsoft: Phi 4 |
|---|---|---|
| Coding agent | $12.00 | $0.74 |
| RAG chat | $16.00 | $0.81 |
| Summarisation (batch) | $14.00 | $0.77 |
| Classification | $10.80 | $0.71 |
| Chat assistant | $26.00 | $0.98 |
| Content generation | $42.00 | $1.26 |
Cite this page
Anthropic: Claude Haiku 4.5 vs Microsoft: Phi 4: $1.000000/$5.000000 vs $0.070000/$0.140000 per 1M tokens, as of 2026-10-04 (Token Barometer, https://tokenbarometer.com/compare/anthropic_claude-haiku-4-5-vs-microsoft_phi-4).
Anthropic: Claude Haiku 4.5 page | Microsoft: Phi 4 page | Open in calculator