Most teams find [Langfuse](https://langfuse.com) when something breaks in production, and they have no idea why. It's the tool that actually shows you what's happening inside your LLM app. Here's what it costs. ## Langfuse Pricing Plans: At a Glance | 🛠 Plan | 💰 Price | 🎯 Best For | 👥 Key Features | |---------|---------|------------|----------------| | **Hobby** | Free | Solo devs, POCs | 50k units/mo, 30-day retention, 2 users | | **Core** | $29/month | Production projects | 100k units/mo, 90-day retention, unlimited users | | **Pro** | $199/month | Scaling teams | 100k units/mo, 3-year retention, high rate limits, SOC2/ISO27001 | | **Enterprise** | $2,499/month | Large orgs | Custom rate limits, SLA, dedicated support engineer | **Overage across all paid plans:** $8 per 100k additional units (volume discounts apply at scale, see pricing calculator below). ## Langfuse Pricing Plans Breakdown **Each plan is built around usage volume,** not headcount. Here's what's behind each plan, what it costs at real usage levels, and who it makes sense for. ### Hobby: Free **What's included:** All platform features (with limits), 50k units/month, 30-day data access, 2 users, community support via GitHub. **Best for:** Solo developers, proof-of-concept projects, and teams evaluating whether Langfuse fits their stack before spending a dollar. #### Pros

✅ Genuinely usable free tier, 50k units is among the most generous in LLM observability

✅ Full feature access (tracing, prompt management, evaluations, SDKs)

✅ No credit card required

#### Cons

❌ Hard limit of 2 users, not viable for teams

❌ 30-day data retention means you lose historical context fast

❌ Rate-limited to 1,000 ingestion requests/min and 30 API requests/min

❌ Community support via GitHub only

### Core: $29/month **What's included:** Everything in Hobby, plus 100k units/month, 90-day data access, unlimited users, in-app support, and 48h response SLO. **Best for:** Small production teams that need unlimited users and longer data retention without breaking the budget. #### Pros

✅ Unlimited users at $29/month flat, no per-seat surprises

✅ Ingestion rate jumps to 4,000 requests/min

✅ Overage is predictable: at 1M units/month total, Core costs ~$101/month ($29 base + $72 overage)

✅ Ingestion rate increases to 4,000 requests/min (vs. 1,000 on Hobby), and general API rate limit rises to 100 requests/min

#### Cons

❌ 90-day retention is decent but won't work for compliance-heavy teams

❌ No SOC2/ISO27001 reports or HIPAA, not suitable for regulated industries

❌ No Enterprise SSO or fine-grained RBAC

### Pro: $199/month **What's included:** Everything in Core, plus 3-year data access, data retention management, unlimited annotation queues, 20,000 req/min ingestion throughput, SOC2 & ISO27001 reports, HIPAA (BAA available), and prioritized in-app support. **Optional Teams add-on:** +$300/month unlocks Enterprise SSO (Okta, Azure AD/EntraID), SSO enforcement, fine-grained project-level RBAC, dedicated Slack channel support (24h SLO), Protected Deployment Labels, and Scheduled Batch Export to Blob Storage. **Best for:** Engineering teams running LLM apps in production who need compliance documentation and serious rate limits. #### Pros

✅ 3-year data history, the only managed tier with this outside Enterprise

✅ SOC2 Type II + ISO27001 + HIPAA included at $199/month

✅ Ingestion throughput jumps to 20,000 req/min, handles high-volume agent pipelines

✅ General API rate limit jumps to 1,000 req/min (vs. 100 on Core), which is ideal for teams running automated eval pipelines

✅ Volume discounts on overages: $7/100k above 1M units, $6.50/100k above 10M, $6/100k above 50M

#### Cons

❌ Teams add-on at +$300/month means full SSO + RBAC costs $499/month total

❌ Protected Deployment Labels and Scheduled Batch Export require the add-on

### Enterprise: $2,499/month **What's included:** Everything in Pro + Teams, plus audit logs, SCIM API, custom rate limits, Uptime SLA, Support SLA, dedicated support engineer, and onboarding/architectural guidance. **Optional with Yearly Commitment:** Custom volume pricing, architecture reviews, AWS Marketplace billing, invoice billing, vendor onboarding, and InfoSec/legal reviews. **Best for:** Large organizations with procurement requirements, strict SLAs, and high enough usage volume to justify custom pricing on a yearly commitment. #### Pros

✅ Custom ingestion rate limits, no ceiling

✅ Formal Uptime SLA and Support SLA

✅ SCIM for automated user provisioning at scale

✅ AWS Marketplace billing (with yearly commitment)

#### Cons

❌ Yearly commitment required for custom volume pricing, AWS Marketplace billing, and InfoSec/legal reviews

❌ High starting price, ROI only makes sense at a serious scale

## Which Langfuse Plan Should You Choose? **Choose Hobby if you:** - Are still figuring out whether Langfuse fits your stack before any money changes hands - Building solo or with one other person, and don't need data older than a month - Just want to see what your LLM app is actually doing without a setup commitment **Choose Core if you:** - Have a small team shipping to production, and per-seat pricing would hurt as you grow - Don't have compliance requirements yet, but need data to stick around longer than 30 days - Want a predictable bill, $29 base plus usage, without enterprise overhead **Choose Pro if you:** - Work in healthcare, finance, or any industry where SOC2 or HIPAA documentation is non-negotiable - Your pipelines are generating enough traffic that 90-day retention isn't cutting it - Need room to add SSO and RBAC later without switching plans **Choose Enterprise if you:** - Your legal or procurement team needs a signed SLA before you can deploy anything - You're running at a scale where custom rate limits and a dedicated support engineer are worth the price - Need to route billing through AWS Marketplace or pay by invoice ## Is Langfuse Worth the Cost? The math works in your favor at most team sizes. Core covers unlimited users at $29/month flat, **so your bill stays the same whether you have 3 engineers or 10.** Pro adds [SOC2 Type II](https://langfuse.com/security/soc2), ISO27001, and HIPAA at $199/month, which **most tools in this category sell as add-ons** or reserve for enterprise contracts. **Langfuse is worth it if you:** - Have a team of 3+ engineers where per-seat pricing would compound fast - Work in a regulated industry and need compliance documentation without moving to an enterprise contract - Use multiple frameworks like LangChain, LlamaIndex, or a custom stack and want one place to observe all of it **Skip Langfuse Cloud if you:** - Are building solo with low volume, the Hobby plan covers you at no cost - Have the infra capacity to self-host, the software is MIT licensed and free, you only pay for your own servers - Are running 10M+ events/month and need the lowest possible per-unit cost. At that scale, Arize Phoenix is worth evaluating - Are building a voice AI product. [**Cekura**](https://www.cekura.ai) **covers what Langfuse doesn't:** concurrent calls, ASR errors, and production alerts. ## Langfuse Pricing Calculator Langfuse uses graduated pricing on overages. **Here's how the tiers break down across all paid plans:** | 📊 Tier | 💰 Rate | |--------|--------| | 0–100k units/mo | Included in base | | 100k–1M units/mo | $8/100k | | 1M–10M units/mo | $7/100k | | 10M–50M units/mo | $6.50/100k | | 50M+ units/mo | $6/100k | **Example:** Core plan at 200k units/month = $29 base + $8 overage = **$37/month total.** **Custom volume pricing is available on Enterprise** with a yearly commitment. ## Langfuse Discounts Langfuse actively supports early teams. **Here's what's available:** - **Early-stage startups:** 50% off, first year - **Research/Students:** Up to 100% off, limits apply - **Open-source projects:** $300 in credits/month, first year ## Langfuse Alternatives & Pricing Comparison As a platform isn't the only option. **Here's how it stacks up against the main alternatives:** | 🛠 Tool | 💰 Starting Price | 🎯 Best For | 🏆 Key Advantage | |--------|-----------------|------------|----------------| | [Langfuse](https://langfuse.com) | [Free or $29/month](https://langfuse.com/pricing) | Teams of any size, multi-framework | Unlimited users, open-source, no per-seat fees | | [LangSmith](https://smith.langchain.com) | [Free or $39/seat/month](https://www.langchain.com/pricing) | LangChain-native teams | Deepest LangChain + LangGraph integration | | [Arize Phoenix](https://phoenix.arize.com) | [Free or usage-based](https://phoenix.arize.com/pricing) | Extreme-scale deployments | Lowest per-unit overage cost at 10M+ events | | [Portkey](https://portkey.ai) | [Free or usage-based](https://portkey.ai/pricing) | LLM cost optimization + routing | AI gateway + observability combo | ## Langfuse + Cekura: Two Tools, One Stack **Langfuse covers your LLM backend well:** what each call costs, where latency spikes, how prompts behave across runs. But it wasn't [built for voice](https://www.cekura.ai/blogs/voice-ai-load-testing-with-cekura-simulate-stress-strengthen), so concurrent call simulation, ASR error detection, and production alerts aren't part of what it does. Cekura was **built for exactly that.** Most teams hit this right before launch, when the LLM backend looks fine but the voice layer hasn't been tested under real load. **Running both** means **you catch that before it reaches users.** **Use Langfuse when you need:** - Deep observability across LangChain, LlamaIndex, OpenAI, or a custom stack - Token and cost tracking without per-seat fees - Full data control via self-hosting (MIT license, zero software cost) **Use Cekura when you need:** - Load testing with 100+ concurrent calls - Real-time alerts for ASR failures and latency spikes in production - HIPAA-compliant voice simulation with human-like scenarios - Predictable pricing starting at $30/month with flexible credits - **Predefined and custom metrics:** built-in latency, instruction-following, and tool-call tracking, plus define your own - **Cron job scheduling:** run tests on a recurring schedule to catch silent regressions between deployments **Use both if** you're shipping a voice AI product to production. Langfuse handles what's happening inside your LLM backend. Cekura handles how your voice layer holds up under real conditions. Together, they cover the full stack. If your stack includes voice, **see how Cekura fits in** [with a 7-day free trial](https://app.cekura.ai/pricing). ## My Bottom Line on Langfuse Pricing Pricing is built around **usage volume,** not headcount. **Core** at $29/month **is the easiest entry point.** Pro at $199/month is worth it if compliance is on your checklist. Enterprise is negotiable on a yearly commitment. Start on Hobby and upgrade to Core once you hit the limits. You'll know within a week if it's the right fit. One thing Langfuse won't tell you is how your [voice layer](https://www.cekura.ai/blogs/performance-testing-voice-agents-practical-guide-cekura) holds up under load, and **that's where Cekura comes in.** ## Frequently Asked Questions ### How Much Does Langfuse Cost? Langfuse costs **$0 on Hobby, $29/month on Core, $199/month on Pro, and $2,499/month on Enterprise.** All paid plans add $8 per 100k units beyond the included limit. ### What Is a Billable Unit? A billable unit is **one trace, observation, or score ingested into the platform.** A single user request that triggers 3 LLM calls and generates 2 evaluation scores counts as 6 units. Multimodal inputs are free while in beta. ### What Happens to My Data if I Downgrade? Data older than your new plan's retention limit gets permanently deleted during the next nightly cleanup. If you drop from Pro (3-year retention) to Core (90-day retention), **anything older than 90 days is gone with no recovery option.** Upgrade before your data expires if you need to keep it. ### Does Langfuse Work With Models Outside OpenAI? Yes, **Langfuse works with Anthropic, Google, Mistral, Ollama, and any model you can instrument via OpenTelemetry or the SDK.** It also auto-calculates costs for most major providers using their published token pricing. ### What's the Difference Between Core and Pro? **Core gives you 90-day retention** and 4,000 req/min ingestion. **Pro extends that to 3 years of retention,** 20,000 req/min, and adds SOC2, ISO27001, and HIPAA. If compliance or data history matters to your team, Pro is the threshold where it becomes available. ### How Do I Test My Voice Agent If I'm Already Using Langfuse? Use **Cekura.** Langfuse traces what's happening inside your LLM calls: token costs, latency, prompt versions. It doesn't cover the voice layer: accent testing, concurrent call simulation, ASR error detection, or drop-off tracking. **Cekura handles that and integrates with your existing Langfuse setup** in under 15 minutes.

Posts tagged with "Ai Tools"

Langfuse Pricing: Plans, Costs & What You Get in 2026