Start

Find your frontier

Automatically evaluate frontier models for your use cases

Get startedSee a sample report
/Support agent/Oct 4 report
Report ready 1 hr after Sonnet 5.5 release
New release

Sonnet 5.5 beats your ticket routing model by 11.9 points at 41% lower cost

Your ticket routing runs GPT-5.6 Sol at $2.90 per task and scores 35.8%
Sonnet 5.5 scores 47.7% at $1.70

Routing score47.7%▲ +33%
Cost per task$1.70▼ −41%
Monthly at 60k tasks$102k▼ −$72k
Ticket routingSwitchSonnet 5.5 Medium−41% cost per task
▲ +11.9 pts
Refund approvalsKeepOpus 5.5 HighNothing beat it under $4
→ 0.0 pts
Reply draftingKeepOpus 5.5 HighSonnet 5.5 Max at 2.4× cost
▼ −0.6 pts
Fable 5.1GPT-5.6 SolGrok 4.7Opus 5.5 MaxSonnet 5.5You todaySwitch here50%40%30%$15$9$3Your frontier26 settings · 6 models

Your intelligent researcher

01

Just describe it

Describe your workflows and Klu shapes the use cases and data for evaluation

02

Pick providers and models

Evaluate today's models, then pick which providers to track as new ones ship

03

Build your data

Let Klu generate your eval set, import it from a data source, or have your own agent send it in

Industrial strength

Learn how teams use Klu.ai for their operations

Tests pass or fail

Engineering

Code review, migration, test generation. Score pass rates on your repo, not SWE-bench.

Query results match

Data analysis

SQL generation, data cleaning, transformations. Scored on whether the result matches, not the code.

Field-level accuracy

Document extraction

Invoices, claims, lab reports into structured fields. The highest-volume LLM job, and the easiest to grade.

Decision accuracy

Operations

Routing, approvals, classification inside a workflow. One decision per task, thousands a day.

Rubric

Customer support

Did the ticket get resolved, and how did it read. At this volume, cost per ticket decides the model.

Misses cost more

Legal and compliance

Contract clauses, policy checks, regulatory flags. No one switches models here without evidence.

Right source, right answer

Knowledge assistants

Q&A over your docs. Every company runs one. Few measure it.

Matches what the rep did

Sales

Lead scoring, reply drafting, call summaries. Grade against what the rep actually did next.

Clinician rubric

Healthcare

Clinical notes, prior auth, coding. Scored by a clinician rubric, with an audit trail for every run.

Expert rubric

Research

Literature synthesis, protein and molecule tasks, protocol drafting. Biology first.

Scaled to your needs

Pro
$50/ month
  • 1,000 credits a month
  • 3 tracked workflows
  • Unlimited runs, models, and test cases
  • Buy credits or auto top-up
  • Reruns on every release
  • Shareable reports
  • Your own API keys, no markup
Start with Pro
Team
$200/ month
  • 6,000 credits a month
  • 15 tracked workflows
  • Unlimited runs, models, and test cases
  • Buy credits or auto top-up
  • Everything in Pro
  • Rubric and custom judges
  • Slack and email on every release
Start with Team
Enterprise
Custom
  • Volume-scaled credits
  • Unlimited tracked workflows
  • Unlimited runs, models, and test cases
  • Usage-based overage billing
  • SSO and audit log
  • Data retention you set
  • Custom terms and support
Talk to us
Enterprise add-onPrivate cloudKlu runs inside your VPC, in your region. API keys, eval data, and reports never leave your account.
Talk to us about private cloud

Encrypted & private

Secure defaults on every plan and deployment

Encrypted before it's stored

Files, extracted text, and provider API keys are locked with AES-256-GCM before database storage

Tamper-evident

Automatically detects unauthorized file alterations and prevents tampered eval data

Role-checked on every request

Enforces workspace roles on every request, so only the right teammates see or edit your use cases

Scoped to its workflow

Serves each file only to the workflow it belongs to, verified by workflow and file ID on every read