Flexible Plans for All Teams
Open Source
Download, install, & run Opik your way
$0
Free
- Full LLM observability & evaluation featureset
- True OSS: same codebase as the hosted versions
Includes:
- LLM tracing & analysis
- Datasets & experiments
- LLM-as-a-judge metrics
- Agent Optimizer Suite
Free Cloud
Perfect for individuals
$0
Free plan
- Unlimited team members
- 25k spans per month
- 60-day data retention
Includes:
- LLM tracing & analysis
- Datasets & experiments
- LLM-as-a-judge metrics
- Agent Optimizer Suite
Pro Cloud
Popular
Expanded usage for teams
$39
Per month
- Unlimited team members
- 100k spans per month
- 60-day data retention
Includes everything in the Free plan plus:
- Customizable monthly span limits
- Customizable data retention periods
Enterprise
Security, compliance & flexible deployments
Custom
- Unlimited team members
- Unlimited spans
Includes everything in the Pro plan plus:
- Flexible deployments
- Service accounts and view-only users
- Single sign-on
- Dedicated support and SLAs
- SOC 2, ISO 27001, ISO 9001, HIPAA, and GDPR compliance
Compare Plans
Observability
| AI Application Tracing Trace every step of your application’s execution path, from context retrieval to model responses to tool calls |  | | | |
| Agent Execution Graphs Visualize your agent’s execution flow | | | | |
| Sessions Track the full session for multi-turn conversations, complex workflows, and conversational agents | | | | |
| Token & Cost Tracking | | | | |
| Multi-Media Logging Log Images, videos, audio files, and more with your traces for future viewing in the UI | | | | |
| Error Surfacing View errors at the project level and use quick-access shortcuts to drill into debugging and error tracking views | | | | |
| User Feedback Tracking Log user feedback with flexible scores and freeform reason explanations using either the UI or the SDK | | | | |
| Integrations with 40+ AI Frameworks, Model Providers, and AI Gateways Opik integrates with dozens of common AI Frameworks and model providers, including LangChain, OpenAI, Google ADK, LangGraph, CrewAI, and many more | | | | |
| OpenTelemetry Integration Opik provides native support for OpenTelemetry (OTel), allowing you to track AI applications in Ruby, Java, and other languages | | | | |
| AI-Powered Debugging with OpikAssist Ask natural language questions to help root-cause issues, identify performance bottlenecks, and get debugging recommendations for your traces | — | | | |
Agent Optimization
| Prompt Optimization Automate prompt engineering and improve your AI application’s performance with Opik’s industry-leading Agent Optimizer toolkit | | | | |
| Tool Optimization Optimize prompts that use external tools and the Model Context Protocol (MCP) | | | | |
| Native Support for 6+ Optimization Algorithms Opik has native support for Evolutionary, Few-Shot Bayesian, MetaPrompt, Hierarchical Reflective Optimizer, MIPRO, and GEPA, with more to come | | | | |
| Optimization Dashboard View your optimization results and performance improvements in Opik’s UI | | | | |
Evaluation
| Agent Evaluation Evaluate the performance of complex AI applications and agent systems – both individual steps and overall task completion | | | | |
| Conversation Evaluation Evaluate full multi-turn conversation as well as individual responses, with built-in and custom metrics | | | | |
| Evaluation Datasets Create and maintain collections of example inputs and expected responses using production traces or synthetic examples for systematic testing and evaluation at scale | | | | |
| Experiments Systematically test your AI application over an entire dataset, using the metrics that matter most to your use case. Use Experiments for benchmarking, A/B testing, and regression testing | | | | |
| Built-in Evaluation Metrics Easy to use, configurable LLM-as-judge and heuristic metrics | | | | |
| Custom Metrics Create custom LLM-as-Judge, criteria-based, and python code based metrics or use metrics from external libraries | | | | |
| Automated Dataset Expansion Automatically expand your dataset for more robust evaluation. Use AI to generate additional synthetic examples to add to your existing dataset | | | | |
| Experiments Dashboard View and comment on experiment results, compare prompts and configurations, and select your best performing version for deployment | | | | |
| Dataset and Evaluation Integrations Ragas, Hugging Face Datasets, Gretel | | | | |
Expert Annotation UI
| Dedicated Annotation UI Opik provides a streamlined, distraction-free interface designed for efficient review by subject matter experts | | | | |
| Custom Feedback Schemas Create custom labels to systematically collect and analyze structured feedback | | | | |
| Annotation Queues Enable subject matter experts to review and annotate agent outputs with easy queues, invitations, and multi-user annotation | | | | |
| Annotation for Traces or Entire Conversations Annotate individual LLM responses or the entire conversation holistically, with purpose-built interfaces for both trace-level and thread-level feedback | | | | |
Prompt Development
| Prompt Library Collaborate on prompt development with both UI-based prompt management and seamless integration with prompts saved in code or files | | | | |
| Prompt Versioning Track and version every change made to a prompt, including edits, timestamps, and authors | | | | |
| Prompt Evaluation Test and compare prompts via either the UI or the SDK by running structured experiments over a dataset | | | | |
| Prompt Playground Test new prompts quickly in Opik’s Prompt playground | | | | |
| Prompt Building Build prompts in the UI using composable user, assistant, and system components. Add variables to easily run evaluations and side-by-side comparisons over a dataset | | | | |
Reliability in Production
| Production-Scale Monitoring Opik has been designed from the ground up to support high volumes of traces making it the ideal tool for monitoring your production LLM applications | | | | |
| Online Evaluation Score your production traces and identify any issues with your production LLM application | | | | |
| AI Guardrails Protect your application by detecting and blocking content from the inputs and outputs of your LLM calls. Opik supports PII, Topic, and Custom guardrails. Available for self-hosted deployment | | — | — | — |
| Alerts Configure automated webhook notifications for important events such as trace errors, new feedback scores, or prompt changes | | | | |
Platform (API, Collaboration, & Data Access)
| Public API Opik’s REST API and complete Python client can be used with both the Open-Source platform and Opik Cloud | | | | |
| API Rate Limits | Unlimited | Unlimited | Unlimited | Unlimited |
| SDKs (Python, TypeScript) | | | | |
| MCP Server Integrate your AI-powered IDE with Opik to use natural language to manage prompts, query traces, access metrics, and more | | | | |
| Query and Export Opik Data Opik’s SDKs and API enable data export and support advanced queries, filtering, and batch operations | | | | |
| UI Data Export Export traces, datasets, experiment data, and annotations to CSV or JSON through the Opik UI | | | | |
| In-Platform Collaboration Collaborate with your team directly in the platform with deep-links to share-able views and commenting on traces, experiments, and prompts | | | | |
| Flexible Deployments Choose from multiple deployment options, including cloud-based, on-premises, and fully managed solutions to fit your infrastructure needs | — | — | — | |
| Optional Add-On: MLOps Platform: Experiment Management, Model Registry, Dataset Management | — | | | |
Security & Compliance
| Data Region | — | US | US | Custom |
| RBAC (Project and Organization Level) | — | — | — | |
| View-Only Users | — | — | — | |
| Enterprise SSO (OAuth 2.0, SAML, and LDAP protocols) | — | — | — | |
| SSO Enforcement | — | — | — | |
| SOC 2, ISO 27001, ISO 9001, HIPAA, and GDPR compliance | — | — | — | |
Support
| Community (Slack; GitHub) | | | | |
| Self-Service (Ask AI; Opik MCP Serve) | | | | |
| Email Support | — | — | | |
| Dedicated Support Team & SLAs | — | — | — | |
| Private Slack Channel | — | — | — | |
| Service Accounts | — | — | — | |
| Response Time SLA | — | — | — | 2 hours |
Monthly Usage
| Spans Included A span represents a single tracked operation within your LLM pipeline, such as a model request, a function call or an agent action | Unlimited | 25k | 100k | Unlimited |
| Additional Spans | Unlimited | — | $5/100k spans | Unlimited |
| Span Data Retention | Unlimited | 60 days | 60 days | Custom |
| Additional Span Retention Increase span data retention from 60 days to 400 days | Unlimited | — | $29/100k spans | Custom |
Questions & Answers
What is the relationship between Comet, Opik, and MLOps / Experiment Management?
Comet is our company name, and we offer two flagship product families: Opik and MLOps. Features, use cases, and pricing models differ between the two products, but they run on the same underlying platform and each comes with access to the free version of the other.
Our MLOps platform is designed for teams building and training machine learning models, with tools for experiment tracking, dataset management, model versioning, and model production monitoring.
Opik is the name of our GenAI observability and evaluation platform, and it’s available in a free open-source version along with the cloud-hosted options you see above. If you are building an application or agent that makes calls to an LLM, use Opik to log, evaluate, iterate, and improve your application’s performance.
Is there a free plan for academics?
Yes, Comet offers a free Pro plan for academic users. Researchers, students, and educators can access the full features of our Pro plan at no cost. To apply for the academic Pro plan, please sign up here and follow the instructions on the Academics page to verify your academic status.
What is the best way to get started with Opik?
The full Opik featureset is free to use for as long as you like, and it’s easy to get started. You can download and self-host the open-source version, or let us handle the infrastructure with the free cloud version. For most individuals and teams in the application testing and development phases, these free plans come with everything you need. You might want to step up to a paid plan as your application or agent usage scales up in production, and/or if you need user management and compliance-related features, but the free versions really do come with everything you need to get up and running.
If you already have an application or agent that makes LLM calls, just follow our Quickstart Guide to integrate Opik with whichever model provider or agent framework you’re already using. And if you’re not ready to log traces yet, we recommend trying out the prompt playground to experiment with user prompts and system prompts and compare models side by side, with no setup steps required.
How can I self-host Opik?
Self-hosting the open-source version of Opik is easy! Just download the code from our GitHub repo and follow the README to get started.
What are the differences between Opik Open Source, Opik Free, Opik Pro, and Opik Enterprise?
The main differences are in usage limits and data retention. All plans include unlimited team members and the full suite of LLM observability and evaluation features. Enterprise plans offer custom hosting and deployment options, advanced identity and authentication management, personalized support plans, increased regulatory compliance, and more.
What is a “span”?
A span is a unit of evaluation or tracing data in Opik, representing a structured input-output pair. This could be an LLM call, a tool call, or any tracked function. A single span can contain many tokens.
Free
Perfect for individuals
$0
Free plan
- 1 platform user
- Generous free tier
Includes:
- Track and compare machine learning training runs
- Dataset management and versioning
- Model Registry
LLM evaluation included for free
Pro
Popular
Advanced collaboration for teams
$39
Per user/month
- Up to 10 users
- 1500 training hours included
Includes everything in the Free plan plus:
- Up to 10 users
- Email support
- Generous storage limits
LLM evaluation included for free
Enterprise
Security, compliance & flexible deployments
Custom
- Unlimited users
- Unlimited training hours
Includes everything in the Pro plan plus:
- Flexible deployments
- Model production monitoring
- Service accounts and view-only users
- Single sign on
- Dedicated support and SLAs
- SOC 2, ISO 27001, ISO 9001, HIPAA, and GDPR compliance
LLM evaluation included for free
Full platform self-hosted
The Comet platform can be deployed locally in just a couple of minutes, perfect if you want a simple single-user deployment.
Comet Installation Options
Self-Serve
Managed by Comet
On-Premise
Compare Plans
Experiment Management
| Experiment tracking Systematically record, compare, and analyze your machine learning training runs to accelerate development and improve model performance. | | | |
| Python visualisations Create tailored visualizations using Python to customize your Comet dashboards, enhancing data insights for your specific needs. | | | |
| Model registry Model Registry centralizes and organizes machine learning models, enabling versioning, tracking, and seamless deployment across your team’s workflow. | | | |
| Dataset management and versioning Track and version training datasets, linking them to experiments for complete model lineage and reproducibility. | | | |
| Hyperparameter search Optimize model performance with automated hyperparameter tuning, tracking all results within the Comet platform. | | | |
| Monthly limits | Fair usage policy | 1500 training hours included $1/training hours | Unlimited |
| Data usage | 100GB | 500GB included $3/100GB/month | Unlimited |
| Team members | 1 | Up to 10 users | Unlimited |
LLM Evaluation
| Tracing Monitor and analyze the performance of your language models during development and in production, capturing inputs, outputs, and metadata for each inference. | | | |
| Datasets Manage and version collections of prompts and examples used for evaluating LLM application performance, ensuring consistent and reproducible assessments. | | | |
| LLM-as-a-judge metrics | | | |
| Production monitoring Track your LLM applications in production with our production ready LLM Evaluation platform. | | | |
| Team members | Unlimited | Unlimited | Unlimited |
Model Production Monitoring
| Data drift detection Automatically identify data drift for your Machine Learning models in production, helping maintain model performance over time. | — | — | |
| Feature distribution analysis Visualize and monitor input feature distributions to detect shifts and anomalies in your data. | — | — | |
| Custom metrics Create and track tailored performance metrics using SQL queries on your logged data. | — | — | |
| Alerts Set up custom notifications for critical events via Slack, email, and other channels. | — | — | |
Platform
| Flexible deployments Choose from multiple deployment options, including cloud-based, on-premises, and fully managed solutions to fit your infrastructure needs. | — | — | |
| Single sign-on Streamline authentication with enterprise-grade SSO, supporting OAuth 2.0, SAML, and LDAP protocols. | — | — | |
| Service accounts Create dedicated machine-to-machine accounts for secure, automated data logging and API access. | — | — | |
| View-only users Grant read-only access to team members for data exploration without the ability to modify or log new data. | — | — | |
| Support Access tiered support options including email, Slack, phone, and dedicated team assistance based on your plan. | Community Slack | Dedicated team |
Questions & Answers
What is the relationship between Comet, Opik, and MLOps / Experiment Management?
Comet is our company name, and we offer two flagship product families: Opik and MLOps. Features, use cases, and pricing models differ between the two products, but they run on the same underlying platform and each comes with access to the free version of the other.
Our MLOps platform is designed for teams building and training machine learning models, with tools for experiment tracking, dataset management, model versioning, and model production monitoring.
Opik is the name of our GenAI observability and evaluation platform, and it’s available in a free open-source version along with the cloud-hosted options you see above. If you are building an application or agent that makes calls to an LLM, use Opik to log, evaluate, iterate, and improve your application’s performance.
Is there a free plan for academics?
Yes, Comet offers a free Pro plan for academic users. Researchers, students, and educators can access the full features of our Pro plan at no cost. To apply for the academic Pro plan, please sign up here and follow the instructions on the Academics page to verify your academic status.
What are the differences between Comet Free, Comet Pro, and Comet Enterprise?
The main differences are in usage limits, features, and support levels:
- Comet MLOps Free: Ideal for individuals and small projects. Includes core experiment tracking features with fair usage limits, 100GB data storage, and community support.
- Comet MLOps Pro: Perfect for growing teams. Offers increased limits (1,500 training hours, 500GB storage), supports up to 10 users, and provides email support. Priced at $39 per user/month.
- Comet MLOps Enterprise: Designed for large organizations. Provides unlimited usage, advanced production monitoring features, flexible deployment options, SSO, dedicated support, and enterprise-grade platform capabilities.
- All plans include essential features like experiment tracking and model registry. Enterprise exclusively offers advanced monitoring, custom metrics, and enhanced security features.
What is a training hour?
A training hour represents 60 minutes of active model training time recorded in Comet. If your model takes 8 hours to train, you’ve used 8 training hours. This time is based on the time between starting and stopping an Experiment.