Claude Haiku 4.5
Claude Haiku 4.5 — Fast, efficient model for latency-sensitive workloads.
- Manufacturer
- Anthropic
- Model ID
claude-haiku-4-5- Supported interfaces
Overview
Claude models use the Anthropic Messages format and are designed for instruction following, tool use, coding, analysis, and extended reasoning according to model tier.
Manufacturer-native capabilities
- Low-latency generation for interactive and high-volume workloads
- Efficient instruction following and extraction
- Streaming and tool use where supported by the manufacturer
The manufacturer source below is authoritative for the complete current set of native capabilities, limits, and modes.
Best use cases
- Coding and technical analysis
- Multi-step tasks and agentic workflows
- Workloads that match the model’s speed and quality tier
Manufacturer-native specification
Claude Haiku 4.5 is positioned by Anthropic for fast interactive coding and high-volume sub-agent work. The values below describe the manufacturer-native model envelope, not a promise that every native platform feature is exposed by every intermediary API.
| Field | Manufacturer specification |
|---|---|
| Context window | 200K tokens |
| Maximum output | 64K tokens |
| Native modalities | Text and image input; text output |
| Native controls | Streaming responses; Tool use and parallel tool workflows; Vision input and prompt caching where enabled |
Manufacturer-native capability highlights:
- Streaming responses
- Tool use and parallel tool workflows
- Vision input and prompt caching where enabled
A native model capability is usable through TokenPass only when it can be represented by one of the interfaces listed at the top of this page. Manufacturer-only hosted tools, beta headers, platform products, and model aliases do not become TokenPass features automatically.
TokenPass integration contract
Send the exact TokenPass model ID claude-haiku-4-5 to https://tokenpass.cloud/v1 using Messages. Streaming is available through the streaming form of the selected interface. Before a production rollout, call GET /v1/models with the target API key and confirm that this exact ID is returned.
Keep request parameters within the selected interface contract. Do not copy a parameter from another manufacturer API merely because the upstream model supports a similar feature. For native limits, modes, and release changes, the official manufacturer source at the end of this page remains authoritative.
Capabilities and limitations
Parameter support depends on the model and listed interface. A manufacturer-native capability does not imply compatibility with another API style. Confirm the callable ID through GET /v1/models.
Manufacturer benchmark highlights
The manufacturer positions this model according to the capability tier described above. See the official source for current results and methodology; TokenPass does not reproduce complete benchmark tables.
Pricing
TokenPass retail pricing is published only in Pricing and Usage.