LangSmith pricing (October 2026 plans): seats, base vs extended traces, LSUs explained
Checked 2026-10-20 on LangChain's own pricing page, usage and billing docs and LangSmith changelog. We don't take payment from LangChain and this page has no affiliate links: how we review.
Short answer: LangSmith charges per seat plus usage. Developer is $0 for one seat with 5,000 base traces a month, Plus is $39 per seat per month with 10,000 base traces, and Enterprise is custom. Since the October 2026 plans, usage is priced in LangChain Standard Units (1 LSU = $1): each base trace past the allowance costs 0.005 LSU, upgrading a trace to 180-day retention adds 0.0025 LSU, and Deployment, Fleet, Engine, Sandboxes and Tuned Evaluators are metered in LSUs on top.
The three plans
Self-serve customers moved to the “October 2026 plans” this month. LangChain's changelog (5-12 Oct 2026) says the new plan applies from the 1st of the month, so the whole month is billed under it; customers with a contract, a credit grant or price adjustments stay on their current plan.
| Developer | Plus | Enterprise | |
|---|---|---|---|
| Price | $0 per seat per month | $39 per seat per month | Custom |
| Seats | 1 (free) | Unlimited, $39 each | Custom |
| Base traces included | 5,000 a month | 10,000 a month | Custom |
| Above the allowance | Pay as you go: 0.005 LSU per base trace | Pay as you go: 0.005 LSU per base trace | Flat 0.0075 LSU per trace unless the contract differs (changelog) |
| Extended (180-day) upgrade | +0.0025 LSU per trace | +0.0025 LSU per trace | No extended surcharge (changelog) |
| Deployment | Not included | 1 free small Serverless deployment, then metered in LSUs | Custom |
| Fleet | 7 LSU included a month | 37 LSU included a month | Custom |
| Engine | Not included | Metered in LSUs | Custom |
| Sandboxes | 8 LSU included; up to 10 sandboxes | 8 LSU included, then metered | Custom |
| Tuned Evaluators | Not included | 0.015 LSU per Perceived Error evaluation run (public beta, US Cloud) | Custom |
| LLM Gateway | Model usage via prepaid Gateway Credits | Gateway controls included; model usage via Gateway Credits | Custom; data protection controls on request |
| Hosting and data location | Cloud | Cloud, LangChain's US or EU region | Cloud, hybrid (self-hosted data plane) or self-hosted in your VPC |
| Workspaces | Not stated | Up to 3 (changelog, 12-19 Oct 2026) | Custom |
| SSO | Not listed | Google, GitHub | Custom SSO |
| Roles | Organization roles (User and Admin) | Organization roles (User and Admin) | Organization roles plus role-based access control (ABAC and RBAC on the plan card) |
| Billing | Monthly, self-serve | Monthly, self-serve | Annual invoice |
What an LSU is
LangChain Standard Units are LangChain's single unit for usage across LangSmith products. One LSU costs $1. How many you use depends on each product's metering rate: a trace, a vCPU-hour of a deployment, an evaluation run, an Engine run.
- On Plus, LSUs are an extra charge on top of the monthly seat fee and any other usage.
- On Enterprise, LSUs are charged against your committed LangChain credits.
- Where to see them: Settings › Billing in LangSmith shows your organization's LSU usage.
- Engine moved to LSU display at the start of October; the changelog (28 Sep-5 Oct 2026) says dollar amounts did not change.
Base vs extended traces, and the 180-day cap
A trace is one execution of your application (an agent run, an evaluator run or a playground session) and can hold many steps. Each trace sits in one of two retention tiers:
| Base | Extended | |
|---|---|---|
| Retention | 14 days | 180 days |
| Price on Developer and Plus | 0.005 LSU per trace above the plan allowance | The base charge plus a 0.0025 LSU upgrade per trace |
| Meant for | Quick debugging and ad-hoc analysis | Traces with feedback from users, evaluators or labelers that you want to keep |
| On your invoice | “LangSmith Traces (Base Charge)” counts all traces, whatever the tier | “LangSmith Traces (Extended Data Retention Upgrades)” counts extended traces |
The 180-day cap: from 14 September 2026, traces ingested into LangSmith SaaS with extended retention are kept for at most 180 days. Older settings above 180 days are applied as 180 days for new traces, new settings above 180 days are rejected, and existing traces are unaffected. Enterprise customers can still set the extended period per workspace up to that maximum. Self-hosted and BYOC deployments are unchanged.
What upgrades a trace to extended automatically
- Online evaluators that score a trace with their retention setting on. Retention extension is on by default for new evaluators; you can opt out per evaluator when the project's default tier is base.
- Automation rules with retention extension on. Matching one run upgrades the whole trace, and a rule on Threads upgrades every trace in the thread.
- Feedback through the API or SDK sent with
extend_trace_retention=true. - Experiments: runs are created at extended retention by default.
- Tuned Evaluators: when one adds feedback to a trace, that trace is upgraded and charged at the standard rate.
- No upgrade: feedback and notes added in the LangSmith UI, and manually adding runs or threads to an annotation queue.
Two details that matter for cost: monitoring charts keep working after base traces expire, because they run on trace metadata, and anything you copy into a dataset is kept indefinitely.
Products metered in LSUs
| Product | Published rate | Included | Notes |
|---|---|---|---|
| Traces (Observability and Evaluation) | 0.005 LSU per base trace; +0.0025 LSU per extended upgrade | 5k (Developer) or 10k (Plus) base traces a month | Enterprise and self-hosted: flat 0.0075 LSU per trace unless the contract differs |
| Deployment | Runtime 0.0675 LSU/vCPU-hr and 0.0090 LSU/GiB-hr; database 0.177 LSU/vCPU-hr and 0.025 LSU/GiB-hr | Plus: 1 small Serverless deployment | Serverless scales to zero when idle; Dedicated is always on |
| Fleet | Metered in LSUs by work done and model tier (Fast, Pro, Max) | 7 LSU (Developer), 37 LSU (Plus) a month | Includes model costs; third-party tool fees are billed by those providers |
| Engine | About 7-45 LSU per run (pricing page estimate) | None | Default org-wide cap 750 LSU a month until an admin changes it |
| Sandboxes | 0.0576 LSU/vCPU-hr, 0.01845 LSU/GiB-hr memory, 0.000123 LSU/GiB-hr storage, billed per second | 8 LSU a month | Developer limited to 10 sandboxes |
| Tuned Evaluators | Perceived Error: 0.015 LSU per evaluation run | None | Public beta on Plus and Cloud Enterprise in the US; only successful runs are billed |
| LLM Gateway | Gateway Credits: prepaid model usage billed as LSUs, tokens at face value | Gateway controls on Plus | 3.5% processing fee on purchased credits; data protection controls are Enterprise-only |
Worked monthly examples
Our arithmetic on LangChain's published rates, not a LangChain quote. Assumptions: list prices, a full month, no taxes or discounts; the extended upgrade fee applies to every upgraded trace, including ones inside the included allowance, because the pricing page lists no included extended traces; no Deployment, Fleet beyond the included LSU, Engine, Sandboxes or Gateway credits.
| Scenario | Plan | Maths | Monthly total |
|---|---|---|---|
| Solo developer, 20,000 traces, all base | Developer (payment method on file) | $0 seat + (20,000 − 5,000) × 0.005 LSU = 75 LSU | $75 |
| Team of 5, 100,000 traces, 10,000 scored by an online evaluator with retention extension on | Plus | 5 × $39 = $195; (100,000 − 10,000) × 0.005 = $450; 10,000 × 0.0025 = $25 | $670 |
| Same team, but the evaluator scores every trace with retention extension on | Plus | $195 + $450 + 100,000 × 0.0025 = $250 | $895 |
| Same team, 100,000 traces, evaluator set to keep traces at base | Plus | $195 + $450 | $645 |
The last three rows show where a LangSmith bill moves: not the seat fee, but how many traces you send and how many end up extended. One deployment adds its own meter: a single vCPU of runtime compute running all month (720 hours) is 0.0675 × 720 = 48.6 LSU ($48.60) before memory and database.
To compare these totals with Langfuse on the same inputs, use the Langfuse vs LangSmith cost calculator; for the other platforms, the pricing calculator.
How to keep the bill in check
- Set the project default tier to base and choose retention per environment; changes apply to new traces only.
- Opt evaluators and automation rules out of retention extension unless you need those traces for 180 days.
- Set usage limits in Settings › Usage and Billing: one for all traces and one for extended traces. They are usage limits, not spend limits, and approximate. When the extended limit is reached, evaluators and rules that would extend retention are blocked.
- Cap Engine spend at organization or project level; LangSmith pauses new Engine runs when the cap is hit.
- Keep what matters in datasets, which have no retention limit, rather than keeping every trace extended.
Limits and discounts worth knowing
- Hourly limits: Developer with no payment method: 50,000 trace events and 500 MB of trace data an hour; Developer with a payment method: 250,000 events and 2.5 GB; Startup/Plus: 500,000 events and 5.0 GB; Enterprise custom. Each trace is capped at 25,000 runs.
- Startups: LangChain lists a Startup Plan with discounted rates and up to $10,000 in credits for VC-backed startups. Per the changelog (12-19 Oct 2026), Startup program customers on Plus move to the October 2026 Startup plans and keep their current term.
- Training on your data: the pricing FAQ says LangSmith does not use your data to train models.
Compare
FAQ
Is LangSmith free?
Yes, for one person. The Developer plan is $0 for 1 seat and includes 5,000 base traces a month; above that you pay as you go. Without a payment method on file, Developer is capped at 5,000 traces a month. Teams need Plus ($39 per seat per month) or Enterprise.
How much does LangSmith cost per trace?
On the October 2026 plans, a base trace above your plan's allowance costs 0.005 LSU, which is $0.005 (about $5 per 1,000). Upgrading a trace to extended (180-day) retention adds 0.0025 LSU. LangChain's changelog says Enterprise and self-hosted traces are billed at a flat 0.0075 LSU per trace with no extended-retention surcharge, unless the contract sets other rates.
What is an LSU in LangSmith?
A LangChain Standard Unit. LangChain calls it the standard unit of measure for usage across all LangSmith products, and 1 LSU = $1. On Plus, LSUs are charged on top of the seat fee; on Enterprise they are drawn from your committed LangChain credits.
What is the difference between base and extended traces?
Retention and price. Base traces are kept 14 days; extended traces are kept 180 days and cost an extra 0.0025 LSU each. Some features upgrade traces to extended automatically, including online evaluators and automation rules with retention extension turned on (the default for new ones).
How much does LangSmith Enterprise cost?
LangChain does not publish an Enterprise price; it is custom, invoiced annually upfront, and includes self-hosted and hybrid options, custom SSO, ABAC and RBAC, a support SLA and custom seats and workspaces.
Does self-hosted LangSmith cost money?
Yes. Self-hosting is an Enterprise option, so it needs an Enterprise agreement. Per the changelog, traces from self-hosted deployments are billed at a flat 0.0075 LSU per trace unless your contract says otherwise.
How is LangSmith Deployment priced?
Plus includes 1 free small Serverless deployment. Further Serverless and Dedicated deployments are metered in LSUs by resources used: runtime compute 0.0675 LSU per vCPU-hour, runtime memory 0.0090 LSU per GiB-hour, database compute 0.177 LSU per vCPU-hour and database memory 0.025 LSU per GiB-hour. Developer has no Deployment.
How is LangSmith Fleet priced?
Fleet is metered in LSUs, and LangChain says the price includes the underlying model costs. Developer includes 7 LSU a month (Fleet pauses when it runs out) and Plus 37 LSU a month, with extra usage billed. Every Fleet run is also traced and counts towards trace billing.
How much does LangSmith Engine cost?
Engine is a Plus and Enterprise feature metered in LSUs. LangChain's pricing page says one Engine run can use roughly 7 to 45 LSU, and the Engine docs set a default organization-wide cap of 750 LSU ($750) a month until an admin changes it.
How are LangSmith Sandboxes priced?
Per second, in LSUs: 0.0576 LSU per vCPU-hour, 0.01845 LSU per GiB-hour of memory and 0.000123 LSU per GiB-hour of storage. Developer and Plus include 8 LSU of sandbox use a month; Developer is limited to 10 sandboxes.
When does LangSmith bill me?
On Developer and Plus, seats are billed monthly on the 1st, pro-rated if you add someone mid-month (no credit for removed seats), and traces are billed monthly in arrears. Enterprise is invoiced annually upfront.
Sources
- LangSmith pricing (plans, per-product rates, FAQ and info notes)
- LangSmith docs: Usage and billing (retention tiers, auto-upgrades, invoice metrics, rate and usage limits)
- LangSmith docs: Engine (manage Engine costs)
- LangSmith Cloud changelog (7-14 Sep, 28 Sep-5 Oct, 5-12 Oct and 12-19 Oct 2026 entries)