Claude Haiku 4.5 API Pricing: Near-Instant Execution with Frontier Precision
Comprehensive Claude Haiku 4.5 API pricing analysis ($1.00/M input, $5.00/M output), prompt caching efficiencies, sub-100ms first-token latency, and coding speed.
Full specs, context window and API limits →How much does Claude Haiku 4.5 cost per million tokens?
Claude Haiku 4.5 costs $1.00 per million input tokens and $5.00 per million output tokens ($2.00/M blended at 3:1). Anthropic fastest model delivering near-instant response times, strong coding proficiency, and lightweight reasoning for high-volume apps. Verified 2026-09-08.
How much does Claude Haiku 4.5 cost per 1,000 requests?
Computed from generated token pricing. Each row assumes the listed input and output tokens per request; this model has no measured verbosity factor, so the unadjusted output estimate is shown.
| Request shape | Input tokens | Output tokens | Cost / 1,000 requests |
|---|---|---|---|
| Short | 100 | 50 | $0.3500 |
| Medium | 1,000 | 500 | $3.5000 |
| Long | 4,000 | 2,000 | $14.0000 |
Formula: ((input price × input tokens) + (output price × output tokens × verbosity factor)) ÷ 1,000,000 × 1,000. Assumptions: short 100/50, medium 1,000/500, long 4,000/2,000 input/output tokens per request. Unadjusted — no measured verbosity factor is available.
Three model-specific pricing decisions
Haiku 4.5 is evaluated for high-volume text workflows. Vision rows remain unit-safe: the registry’s text rate is not converted into an image price.
1. Classification, chat, and vision-shaped monthly bills
| Workload | Fixed shape | 100K requests | Unit treatment |
|---|---|---|---|
| Classification | 300 / 30 tokens | $45.00 | Text tokens |
| Chat | 1,000 / 500 tokens | $350.00 | Text tokens |
| Vision-shaped | 1,000 text + image | $350.00 | Image price unavailable; text proxy is not an image bill |
2. Haiku → Sonnet 4.6 accepted-result and retry crossover
| Fixed workload (input / output) | Claude Haiku 4.5 | Claude Sonnet 4.6 | Decision boundary |
|---|---|---|---|
| Short prompt · 1,000 / 500 | $350.00 | $1050.00 | 10% accepted-result uplift required to justify the higher bill |
| Long prompt · 8,000 / 800 | $1200.00 | $3600.00 | 15% accepted-result uplift required to justify the higher bill |
Formula: 100,000 × (input tokens × input $/M + output tokens × output $/M) ÷ 1,000,000. Uplift is a decision threshold, not a measured quality claim.
3. Cache-write/read and batch reuse break-even
| Mechanic | Dated rate evidence | Break-even / result |
|---|---|---|
| Cache write | Provider/model-specific multiplier unavailable | Unavailable — do not assume a discount |
| Cache read | Provider/model-specific multiplier unavailable | Unavailable — do not assume free input |
| Batch | Model-specific batch rate unavailable | Unavailable — turnaround and discount not inferred |
| TTL | Model-specific TTL unavailable | Reuse horizon is user-supplied |
Verified 2026-04-06. Luna is the data owner for this server-rendered module. “Unavailable” means no compatible dated evidence was found; it is never treated as zero. First-party price source · Run this exact scenario.
All three decisions are server-rendered for Claude Haiku 4.5; fixed inputs, formulas, dated sources, and unavailable states are intentionally visible.
Batch 62 · exact-model pricing decision contributions · verified 2026-09-07
Exact model boundary: Anthropic Claude Haiku 4.5 (claude-haiku-4-5). Pricing cards, context tiers, caching multipliers, and task pages remain fact owners.
Prompt caching read/write economics on fast agents
Frozen Batch 62 scenario board. Formula / deterministic rule: cost = (uncached_in * 1.00 + cache_write * 1.25 + cache_read * 0.10 + out * 5.00) / 1M Boundary: Owns Haiku 4.5 prompt caching savings and latency benefits.
| Frozen scenario / field ID | Exact identity and evidence fields | Result | State |
|---|---|---|---|
batch62-claude-haiku-4-5-m1-r1interactive uncached query | model=claude-haiku-4-5; provider=Anthropic; slug=claude-haiku-4-5; scenario=interactive uncached query; interactive turns; prompt tokens; cached tokens; output tokens; net cost; effective token rate; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-07; measurement versus assumption=explicit | Unavailable — interactive uncached query is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
batch62-claude-haiku-4-5-m1-r23-step rapid classification flow | model=claude-haiku-4-5; provider=Anthropic; slug=claude-haiku-4-5; scenario=3-step rapid classification flow; interactive turns; prompt tokens; cached tokens; output tokens; net cost; effective token rate; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-07; measurement versus assumption=explicit | Unavailable — 3-step rapid classification flow is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
batch62-claude-haiku-4-5-m1-r310-step lightweight agent loop | model=claude-haiku-4-5; provider=Anthropic; slug=claude-haiku-4-5; scenario=10-step lightweight agent loop; interactive turns; prompt tokens; cached tokens; output tokens; net cost; effective token rate; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-07; measurement versus assumption=explicit | Unavailable — 10-step lightweight agent loop is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
batch62-claude-haiku-4-5-m1-r450-step high-frequency chat stream | model=claude-haiku-4-5; provider=Anthropic; slug=claude-haiku-4-5; scenario=50-step high-frequency chat stream; interactive turns; prompt tokens; cached tokens; output tokens; net cost; effective token rate; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-07; measurement versus assumption=explicit | Unavailable — 50-step high-frequency chat stream is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
batch62-claude-haiku-4-5-m1-r5cache invalidation on cold restart | model=claude-haiku-4-5; provider=Anthropic; slug=claude-haiku-4-5; scenario=cache invalidation on cold restart; interactive turns; prompt tokens; cached tokens; output tokens; net cost; effective token rate; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-07; measurement versus assumption=explicit | Unavailable — cache invalidation on cold restart is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
batch62-claude-haiku-4-5-m1-r6unregistered cache breakpoint | model=claude-haiku-4-5; provider=Anthropic; slug=claude-haiku-4-5; scenario=unregistered cache breakpoint; interactive turns; prompt tokens; cached tokens; output tokens; net cost; effective token rate; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-07; measurement versus assumption=explicit | Unavailable — unregistered cache breakpoint is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
First-party provenance: Anthropic official API pricing; verification date 2026-09-07. Missing or conflicting joins fail closed.
High-throughput real-time classification and filtering
Frozen Batch 62 scenario board. Formula / deterministic rule: monthly_spend = calls * ((in_tokens * 1.00 + out_tokens * 5.00) / 1M) Boundary: Owns high-volume classification and sentiment analysis budgeting for Haiku 4.5.
| Frozen scenario / field ID | Exact identity and evidence fields | Result | State |
|---|---|---|---|
batch62-claude-haiku-4-5-m2-r1100K customer message classifications | model=claude-haiku-4-5; provider=Anthropic; slug=claude-haiku-4-5; scenario=100K customer message classifications; query volume; prompt tokens; completion tokens; monthly spend; cost per 10K queries; SLA status; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-07; measurement versus assumption=explicit | Unavailable — 100K customer message classifications is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
batch62-claude-haiku-4-5-m2-r2500K real-time moderation events | model=claude-haiku-4-5; provider=Anthropic; slug=claude-haiku-4-5; scenario=500K real-time moderation events; query volume; prompt tokens; completion tokens; monthly spend; cost per 10K queries; SLA status; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-07; measurement versus assumption=explicit | Unavailable — 500K real-time moderation events is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
batch62-claude-haiku-4-5-m2-r31M streaming chat response queries | model=claude-haiku-4-5; provider=Anthropic; slug=claude-haiku-4-5; scenario=1M streaming chat response queries; query volume; prompt tokens; completion tokens; monthly spend; cost per 10K queries; SLA status; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-07; measurement versus assumption=explicit | Unavailable — 1M streaming chat response queries is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
batch62-claude-haiku-4-5-m2-r45M webhook automated triage records | model=claude-haiku-4-5; provider=Anthropic; slug=claude-haiku-4-5; scenario=5M webhook automated triage records; query volume; prompt tokens; completion tokens; monthly spend; cost per 10K queries; SLA status; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-07; measurement versus assumption=explicit | Unavailable — 5M webhook automated triage records is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
batch62-claude-haiku-4-5-m2-r5rate limit quota exhaustion buffer | model=claude-haiku-4-5; provider=Anthropic; slug=claude-haiku-4-5; scenario=rate limit quota exhaustion buffer; query volume; prompt tokens; completion tokens; monthly spend; cost per 10K queries; SLA status; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-07; measurement versus assumption=explicit | Unavailable — rate limit quota exhaustion buffer has no matched, dated bilateral observation. | FAIL CLOSED — manual, probe, or source evidence required |
batch62-claude-haiku-4-5-m2-r6unhandled classification schema error | model=claude-haiku-4-5; provider=Anthropic; slug=claude-haiku-4-5; scenario=unhandled classification schema error; query volume; prompt tokens; completion tokens; monthly spend; cost per 10K queries; SLA status; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-07; measurement versus assumption=explicit | Unavailable — unhandled classification schema error is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
First-party provenance: Anthropic Claude documentation; verification date 2026-09-07. Missing or conflicting joins fail closed.
Haiku 4.5 vs Sonnet 4.6 tier escalation economics
Frozen Batch 62 scenario board. Formula / deterministic rule: router_cost = haiku_pct * haiku_cost + sonnet_pct * sonnet_cost Boundary: Owns tiered model escalation routing between Haiku 4.5 and Sonnet 4.6.
| Frozen scenario / field ID | Exact identity and evidence fields | Result | State |
|---|---|---|---|
batch62-claude-haiku-4-5-m3-r1100% Haiku 4.5 baseline | model=claude-haiku-4-5; provider=Anthropic; slug=claude-haiku-4-5; scenario=100% Haiku 4.5 baseline; escalation mix; monthly requests; blended spend; savings vs pure Sonnet; quality trade-off; router verdict; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-07; measurement versus assumption=explicit | Unavailable — 100% Haiku 4.5 baseline is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
batch62-claude-haiku-4-5-m3-r290% Haiku / 10% Sonnet 4.6 escalation | model=claude-haiku-4-5; provider=Anthropic; slug=claude-haiku-4-5; scenario=90% Haiku / 10% Sonnet 4.6 escalation; escalation mix; monthly requests; blended spend; savings vs pure Sonnet; quality trade-off; router verdict; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-07; measurement versus assumption=explicit | Unavailable — 90% Haiku / 10% Sonnet 4.6 escalation is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
batch62-claude-haiku-4-5-m3-r375% Haiku / 25% Sonnet 4.6 escalation | model=claude-haiku-4-5; provider=Anthropic; slug=claude-haiku-4-5; scenario=75% Haiku / 25% Sonnet 4.6 escalation; escalation mix; monthly requests; blended spend; savings vs pure Sonnet; quality trade-off; router verdict; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-07; measurement versus assumption=explicit | Unavailable — 75% Haiku / 25% Sonnet 4.6 escalation is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
batch62-claude-haiku-4-5-m3-r450% Haiku / 50% Sonnet 4.6 escalation | model=claude-haiku-4-5; provider=Anthropic; slug=claude-haiku-4-5; scenario=50% Haiku / 50% Sonnet 4.6 escalation; escalation mix; monthly requests; blended spend; savings vs pure Sonnet; quality trade-off; router verdict; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-07; measurement versus assumption=explicit | Unavailable — 50% Haiku / 50% Sonnet 4.6 escalation is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
batch62-claude-haiku-4-5-m3-r5100% Sonnet 4.6 comparison | model=claude-haiku-4-5; provider=Anthropic; slug=claude-haiku-4-5; scenario=100% Sonnet 4.6 comparison; escalation mix; monthly requests; blended spend; savings vs pure Sonnet; quality trade-off; router verdict; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-07; measurement versus assumption=explicit | Unavailable — 100% Sonnet 4.6 comparison is a frozen fixture pending exact identity, configuration, and denominator joins. | UNTESTED — assumption cannot establish a verdict |
batch62-claude-haiku-4-5-m3-r6unresolved classifier confidence score | model=claude-haiku-4-5; provider=Anthropic; slug=claude-haiku-4-5; scenario=unresolved classifier confidence score; escalation mix; monthly requests; blended spend; savings vs pure Sonnet; quality trade-off; router verdict; prompt/config/input/output/result/cache/checkpoint/artifact hashes=required; evidence=2026-09-07; measurement versus assumption=explicit | Unavailable — unresolved classifier confidence score has no matched, dated bilateral observation. | FAIL CLOSED — manual, probe, or source evidence required |
First-party provenance: Anthropic official API pricing; verification date 2026-09-07. Missing or conflicting joins fail closed.
Method and limitations: formulas are deterministic; observed and assumed inputs are labeled; no missing provider, host, account, region, realm, alias, snapshot, revision, weight, artifact, control, tool, modality, workload, rate-period, timestamp, or result is transferred. Run the Claude Haiku 4.5 Batch 62 scenario →
claude-haiku-4-5Claude Haiku 4.5 API Pricing: Near-Instant Execution with Frontier Precision
Claude Haiku 4.5 costs $1.00 per million input tokens and $5.00 per million output tokens ($2.00/M blended at 3:1). Anthropic fastest model delivering near-instant response times, strong coding proficiency, and lightweight reasoning for high-volume apps. Verified 2026-09-08.
Claude Haiku 4.5 combines blazing inference speed with an accessible $2.00/M blended price.
| Scenario | Rendered Evidence & Bounds |
|---|---|
| Scenario 1 | Rapid syntax error diagnostic check (1K in, 200 out): $0.002000 per diagnostic run |
| Scenario 2 | Customer support chat triage turn (1.5K in, 300 out): $0.003000 per dialogue turn |
| Scenario 3 | Code snippet autocomplete expansion (800 in, 150 out): $0.001550 per keystroke call |
| Scenario 4 | Structured JSON classification turn (2K in, 250 out): $0.003250 per classified item |
| Scenario 5 | Pull request commit message generator (4K in, 200 out): $0.005000 per commit review |
| Scenario 6 | Monthly 50M token high-speed production fleet: $100.00 infrastructure budget |
Automatic prompt caching further compresses input costs for latency-critical customer agents.
| Scenario | Rendered Evidence & Bounds |
|---|---|
| Scenario 1 | Shared application system prompt cache (8K tokens): 80% input cost reduction |
| Scenario 2 | Interactive IDE developer session (10 turns cached): 81% cumulative input savings |
| Scenario 3 | Customer chat history cache across 8 turns: 78% net input cost reduction |
| Scenario 4 | Cache write fee ($1.25/M) amortized after only 1.25 repeat queries within 5 minutes |
| Scenario 5 | Near-zero cache retrieval latency: slashes TTFT by 50% for fluid conversational apps |
| Scenario 6 | Net operational cost drops below $1.20/M blended for high-frequency interactive tools |
Tiered routing between Haiku 4.5 and Sonnet 5 delivers 42.5% cost savings and 65% faster speed.
| Scenario | Rendered Evidence & Bounds |
|---|---|
| Scenario 1 | 1M developer queries routed via tiered architecture: $2,300.00 vs $4,000.00 monolithic Sonnet 5 |
| Scenario 2 | Haiku 4.5 handles 85% routine syntax fixes, documentation lookups, and unit tests |
| Scenario 3 | Sonnet 5 ($2.00/$10.00) reserved for complex multi-file architectural refactors |
| Scenario 4 | Developer perceived response speed improves by 65% due to Haiku sub-second generation |
| Scenario 5 | Net engineering infrastructure spend reduced by 42.5% across the organization |
| Scenario 6 | Uniform Anthropic API schemas allow seamless handoff between model tiers |
How fast is Claude Haiku 4.5?
How much does Claude Haiku 4.5 cost at scale?
| Tokens / month | Est. cost (blended 3:1) |
|---|---|
| 100,000 | $0.20 |
| 1,000,000 | $2.00 |
| 10,000,000 | $20.00 |
| 100,000,000 | $200.00 |
How does Claude Haiku 4.5 compare with other models?
What is Claude Haiku 4.5 best for?
What should you explore next for Claude Haiku 4.5?
Which Claude Haiku 4.5 head-to-head comparisons are available?
What are common questions about Claude Haiku 4.5?
Is Claude Haiku 4.5 cheaper than Muse Spark 1.3?
Claude Haiku 4.5 costs $2.00/M blended tokens, Muse Spark 1.3 costs $2.00/M — Muse Spark 1.3 is cheaper.
How much does 1 million tokens cost with Claude Haiku 4.5?
At a 3:1 input:output ratio, 1 million blended tokens costs approximately $2.00. Pure input costs $1.00/M; pure output costs $5.00/M.
What does Claude Haiku 4.5 cost at high volume?
At 100 million blended tokens a month, Claude Haiku 4.5 costs approximately $200.00. See the cost-at-scale table below for other volumes.