Anthropic
Claude Haiku 4.5 (Thinking)
Claude Haiku 4.5 (Thinking) is a model from Anthropic. On the Vals AI Legal Research Bench it reaches 10.58 percent all-pass, position 80 of 91, within a standard error of ±2.14 percentage points at $0.27 per task, as at 9 November 2026 [1]. It passes the compliance gate and reaches a PrudentScore of 32 points.
- Generally available
- 15 October 2025
- PrudentScore
- 32out of 100position 26 of 30 with a score
- Compliance gate
- PassesGoogle Cloud Vertex AI: EU multi-region (ML processing in the EU) per Google's data-residency table (updated 2026-11-10); Google Cloud EMEA Limited (Dublin) is the contracting party for EMEA customers except France, Italy and Poland; Service Specific Terms §18 Training Restriction: "Google will not use Customer Data to train or fine-tune any AI/ML models without Customer's prior permission or instruction"
- Cost per task
- $0.27Vals AI, per completed task [1]
01
Compliance gate
Why
- Google Cloud Vertex AI: EU multi-region (ML processing in the EU) per Google's data-residency table (updated 2026-11-10); Google Cloud EMEA Limited (Dublin) is the contracting party for EMEA customers except France, Italy and Poland; Service Specific Terms §18 Training Restriction: "Google will not use Customer Data to train or fine-tune any AI/ML models without Customer's prior permission or instruction".
- AWS Bedrock: In-Region in EU regions: eu-north-1, eu-west-1 (and EU Geo cross-region) per AWS's regional availability table; Contracting party Amazon Web Services EMEA SARL (Luxembourg) for customers with an EMEA account [AWS GDPR whitepaper]; AWS Bedrock FAQ: "your content is not used to improve the base models and is not shared with any model providers". AWS is not a current Prudai processor.
The gate only opens when all three requirements are covered: processing in the EU, a contracting party in the EU or EEA, and a contractual exclusion of training on customer data. The gate tests a route, not a model name: the same model can pass through one provider and fail through another. Where a provider publishes no figure, we give no estimate.
Sources for this verdict
- docs.cloud.google.com/gemini-enterprise-agent-platform/resources/data-residency
- cloud.google.com/terms/google-entity
- cloud.google.com/terms/service-terms
- docs.aws.amazon.com/bedrock/latest/userguide/models-region-compatibility.html
- docs.aws.amazon.com/whitepapers/latest/navigating-gdpr-compliance/in-detail.html
- aws.amazon.com/bedrock/faqs/
02
Routes and EU processing
The gate tests a route (processor, region and contract), not a model name. Per route: processing in the EU, the contracting party and training on customer data.
| Processor | Region or deployment | EU processing | Source |
|---|---|---|---|
| Microsoft Azure (Foundry Models) | Global Standard (hosted by Azure) and Data Zone Standard in the US only, no EU Data Zone Standard | No | SourceRead 10 November 2026 |
| Google Cloud Vertex AI | EU multi-region (ML processing in the EU) | Yes | SourceRead 10 November 2026 |
| AWS Bedrock | In-Region in EU regions: eu-north-1, eu-west-1 (and EU Geo cross-region) | Yes | SourceRead 10 November 2026 |
03
Scores per benchmark
Every measurement with its source, measurement date and the day we read it. The position counts only the models measured on the same benchmark and version.
- Legal Research BenchVersion 1 · Counts towards the PrudentScore10.58%± 2.1480 of 91
Vals AI, Legal Research Bench, v1, 9 November 2026 · Source · Measured 9 November 2026 · Read 10 November 2026 · cost per task $0.27
- Harvey LAB (Vals implementation)Version 1 · Counts towards the PrudentScore0.83%± 0.8360 of 80
Vals AI, Harvey LAB (Vals implementation), v1, 18 October 2026 · Source · Measured 18 October 2026 · Read 10 November 2026 · cost per task $0.42
- LegalBench (Vals leaderboard)Version 181.24%± 0.4977 of 113
Vals AI, LegalBench (Vals leaderboard), v1, 13 October 2026 · Source · Measured 13 October 2026 · Read 10 November 2026
- CaseLaw v2Version 256.48%± 0.9230 of 54
Vals AI, CaseLaw v2, v2, 4 May 2026 · Source · Measured 4 May 2026 · Read 10 November 2026
- TaxEval v2Version 267.54%± 0.9189 of 108
Vals AI, TaxEval v2, v2, 1 September 2026 · Source · Measured 1 September 2026 · Read 10 November 2026
- Tax Agent BenchVersion 15.86%± 1.5477 of 82
Vals AI, Tax Agent Bench, v1, 9 November 2026 · Source · Measured 9 November 2026 · Read 10 November 2026 · cost per task $0.23
- CorpFin v2Version 260.61%± 0.9667 of 108
Vals AI, CorpFin v2, v2, 12 August 2026 · Source · Measured 12 August 2026 · Read 10 November 2026
- Immigration BenchNot measured
- Access to JusticeNot measured
- Public Benefits BenchNot measured
- Vals Legal IndexNot measured
- Vals IndexNot measured
- GPQA Diamond (Vals-run)Version 172.22%± 2.4783 of 109
Vals AI, GPQA Diamond (Vals-run), v1, 1 September 2026 · Source · Measured 1 September 2026 · Read 10 November 2026
- MMLU Pro (Vals-run)Version 178.72%± 0.4892 of 110
Vals AI, MMLU Pro (Vals-run), v1, 1 September 2026 · Source · Measured 1 September 2026 · Read 10 November 2026
- Artificial Analysis Intelligence IndexNot measured
Source: Vals AI (Legal Research Bench, Harvey LAB (Vals implementation), LegalBench (Vals leaderboard), CaseLaw v2, …) · Licence: CC BY 4.0 · Adaptation: Prudai selected the models, translated the explanations, added the gate verdict and the PrudentScore and rounded figures to two decimals.
The PrudentScore is a composition by Prudai B.V. based on publicly published figures and does not represent the view or endorsement of Vals AI, Harvey or Artificial Analysis. Figures from Artificial Analysis appear in running text only, with attribution and a link, as that party's terms require [9].
04
Key facts
- Developer
- Anthropic
- Model family
- Claude
- Release date
- 15 October 2025
- Status
- Generally available
- Context window
- 200K tokens
- Max output tokens
- 64K tokens
- List price
- $1.00 input · $5.00 outputper million tokens · Read 10 November 2026
- Weights
- Closed
- = not yet recorded in the register.
05
Frequently asked questions about this model
Does Claude Haiku 4.5 (Thinking) pass the compliance gate for legal work?
Yes. Google Cloud Vertex AI: EU multi-region (ML processing in the EU) per Google's data-residency table (updated 2026-11-10); Google Cloud EMEA Limited (Dublin) is the contracting party for EMEA customers except France, Italy and Poland; Service Specific Terms §18 Training Restriction: "Google will not use Customer Data to train or fine-tune any AI/ML models without Customer's prior permission or instruction". AWS Bedrock: In-Region in EU regions: eu-north-1, eu-west-1 (and EU Geo cross-region) per AWS's regional availability table; Contracting party Amazon Web Services EMEA SARL (Luxembourg) for customers with an EMEA account [AWS GDPR whitepaper]; AWS Bedrock FAQ: "your content is not used to improve the base models and is not shared with any model providers". AWS is not a current Prudai processor. Route: Google Cloud Vertex AI. The gate tests a route, not a model name; the sources are on this page.
What does Claude Haiku 4.5 (Thinking) cost per task?
Vals computes $0.27 per completed task on the Legal Research Bench [1]. The list price is $1.00 per million input tokens and $5.00 per million output tokens.
How does Claude Haiku 4.5 (Thinking) score on legal research?
10.58 percent all-pass on the Vals AI Legal Research Bench, position 80 of 91 measured models, as at 9 November 2026 [1]. That benchmark measures US law; an independent, published benchmark for Dutch legal work does not exist yet.
Related models
Claude and Anthropic, GPT and OpenAI, Gemini and Google, Grok and xAI, Kimi and Moonshot AI, Muse and Meta, Mistral, DeepSeek, GLM and Z.ai, Harvey, Vals AI and Artificial Analysis are trademarks of their respective owners. They are named for identification only and this implies no partnership, sponsorship or endorsement.
Brand marks appear for identification only and imply no partnership, sponsorship or endorsement.