Claude Haiku 4.5

Optimized for speed: Claude Haiku 4.5 is Anthropic’s compact model for real-time interactions and high-throughput API pipelines. It processes text and image inputs with a context window of 200,000 tokens, supports tool calls, and delivers responses at a price of $1 per million input tokens and $5 per million output tokens. Anthropic has announced retirement for October 15, 2026.

Anthropic Version 4.5 Commercial use permitted Dense 200 K Context 01/2025 $1 / $5 per 1M

Proprietary
Frontier
Anthropic
Text
Vision
Instruction-Tuned
Real-Time

Sovereign Risk: MEDIUM Anthropic is a US-based company and subject to the CLOUD Act; the model weights are not publicly accessible.

Tool-use details

Asset performance, reliability, and runtime profile

Name	P1	P2	Combined
EU License Research	100	40	64
HTTP Fetch & Extract	80	80	80
Tool Failure Handling (404)	40	40	37
Web Search & Tool Selection	100	80	91
URL Construction & Fetch	80	35	57.5
Multilingual Search & Synthesis	100	40	70

Name

Combined

EU License Research

100

HTTP Fetch & Extract

Tool Failure Handling (404)

Web Search & Tool Selection

100

URL Construction & Fetch

57.5

Multilingual Search & Synthesis

100

CrucibleMark evaluates tool use across 6 independent tests. Click a test name for details.

Reliability

Tool Call Valid: No

Retry: Not required

Hallucination: Detected

Reliability measures how consistently a model actually executes tool calls: Tool Call Valid schema and format accepted, Retry Required successful only after retry, Hallucination Flag fabricated tools or parameters detected. All three green means production-ready.

Runtime profile

Call 1: 5.26; First Request

MCP: 1.18; Protocol Latency

Synthesis: 3.67; Response Generation

Total: 60.68; Sum of All Phases

Token: 12593; Input + Output

Cost: $0; Cost per Run

The runtime profile shows the latency and cost metrics for the model run: Call 1 First Request, MCP Protocol Latency, Synthesis Response Generation, Total sum of all phases, supplemented by Token input and output and Cost cost per run.

Deployment Verdict

Conditional deploy, because tool usage is operationally functional in most cases, but an invalid tool call and the hallucination signal — at only moderate overall performance — limit confidence in a production MCP pipeline.

Tool Execution Profile

Claude Haiku 4.5 selects tools intelligently rather than purely by rote. In the Web Search & Tool Selection test, which checks whether the model distinguishes between search and direct retrieval without an explicit hint, it correctly identifies the need for web_search. This speaks to usable tool selection in open-ended tasks. In the URL Construction test, which measures the autonomous derivation of a target URL followed by a fetch, it performs only partially well. URL construction is workable, but not deterministic enough for pipelines where the first call must land correctly. The critical finding is that the tool call was not consistently valid overall. This is not a retry issue but a protocol and execution risk: the model apparently understands the tool schema in principle, yet does not produce it stably enough for unattended handoff.

Synthesis Fidelity

How well does it condense tool results? Only with limited reliability. Condensation quality is visibly below execution quality. This is evident in EU License Research, URL Construction & Fetch, and Multilingual Search & Synthesis: the model often retrieves the information but compresses it imprecisely, omits relevant details, or draws conclusions that are too weak. For production tool pipelines this is problematic, because the value of the infrastructure only materializes in clean processing of the results.

Does it stay within the tool result or fall back on training data? In the honeypot EU License Research test — which checks whether current license restrictions are answered from web sources rather than training knowledge — it does not hallucinate. That is the most important trust signal. At the same time, a hallucination signal appears in the findings at the run level. This is not merely a quality deficiency but a safety risk: once a model outputs fabricated facts as a tool result, it undermines the reliability of the entire pipeline.

Error Resilience

In the 404 test, which checks for transparent behavior when a tool call fails, Claude Haiku 4.5 does not fabricate page content. That is the minimum requirement for production and is met here. However, communication of the failure remains only moderately clear. Acceptable for supervised workflows, but not strong enough for autonomous chains with hard downstream decisions.

Operational Profile

Call 1: 5.26s. MCP latency: 1.18s. Call 2: 3.67s. Total: 60.68s.
Price: $1.0/M input, $5.0/M output.
Direct assessment: an inexpensive model, but the end-to-end runtime is not convincingly fast given the reliability delivered.

Conclusion & Recommendation

Suitable for cost-conscious, supervised assistance and research pipelines where tool selection matters more than precise final synthesis and a human reviews the output. Not suitable for compliance, retrieval, or decision workflows that require protocol-stable tool calls and strict fidelity to tool results. If you deploy it, do so with strict schema validation, output checks, and a second verification stage after every critical tool result.

This analysis was generated automatically based on benchmark data. Model used: GPT 4.5 by OpenAI. The raw data and full methodology are documented in the GitHub project.