Browserbase vs Browserless vs Steel: Real AI Browser Automation Cost per 10,000 Browser Hours in 2026
The short answer: at 10,000 browser-hours a month with the concurrency this workload actually needs (a peak of roughly 40 simultaneous browsers), Steel's Scale plan is the cheapest for browser compute alone at about $950, Browserbase's Startup plan costs about $1,049, and Browserless's unit-based Scale plan costs about $1,400 once its 500,000 included units are exceeded. Steel's cheaper Launch plan ($0.10/hour) only supports 10 concurrent sessions, well under this workload's peak requirement, so Scale ($250 base, $0.08/hour, 100 concurrent) is the realistic comparison point, not Launch. Add an illustrative 1,250 GB of residential proxy traffic and proxy bandwidth takes over the bill on every vendor that discloses a rate for it: Steel Scale rises to about $8,450 (89% proxy) and Browserbase Startup to about $13,499 (92% proxy). Browserless now publishes enough information to price browser time precisely — its $0.0015-per-unit overage rate and 30-second unit definition make the $1,400 compute figure a real number, not an unknown — but its official rules for how proxy traffic itself consumes units are not detailed enough in its public pricing to extend that calculation into the proxy-heavy scenario.
The billing-unit mismatch, up front
Browserbase bills browser hours (with a reported one-minute minimum per session), proxy gigabytes, and separate Search, Fetch and Extract call counts. Steel bills in dollar credits that convert to browser hours, proxy gigabytes and CAPTCHA solves at plan-specific rates, with its two self-serve tiers (Launch and Scale) priced and capacity-gated differently enough that they are not interchangeable at this workload's concurrency. Browserless bills units, where one unit is 30 seconds of browser connection time, with a published overage rate once a plan's included units are exhausted; proxy traffic is stated to also consume units, but the exact consumption rule for proxy bandwidth specifically is not detailed in Browserless's public pricing materials. Normalizing therefore means converting browser-hours to Browserless units (1 hour = 120 units), tracking proxy gigabytes separately for every vendor because proxy traffic is metered separately on all three, and treating concurrency as a plan gate that can rule out an otherwise-cheaper tier entirely.
What each vendor actually bills
Browserbase (OFFICIAL, docs.browserbase.com/guides/plans-and-pricing). Free: $0, 1 browser hour, 15-minute sessions. Developer: $20 a month, 100 browser hours, then $0.12 an hour; 1 GB of proxy, then $12 per GB; 25 concurrent browsers; 6-hour sessions. Startup: $99 a month, 500 browser hours, then $0.10 an hour; 5 GB of proxy, then $10 per GB; 100 concurrent browsers; basic CAPTCHA solving; 30-day retention. Scale: custom, 250+ concurrency. Exceeding included usage on Developer or Startup is billed at overage rates rather than cut off. Search ($7 per 1,000 beyond 1,000), Fetch ($0.50–$1 per 1,000, $4 with proxies) and Extract calls are separate meters that widen the bill beyond browser hours.
Steel (OFFICIAL, docs.steel.dev/overview/pricing). Launch: $0 base plus usage, a one-time $30 signup usage credit (not a recurring monthly allowance), $0.10 per browser-hour, $10 per proxy GB, 10 concurrent sessions. Scale: $250 a month plus usage, including $100 of usage credit applied monthly against that usage, $0.08 per browser-hour, $6 per proxy GB, 100 concurrent sessions. Enterprise: custom. Steel's two self-serve tiers differ in concurrency ceiling as much as in rate — Launch's 10-concurrent limit rules it out for any workload needing more simultaneous sessions than that, regardless of how cheap its per-hour rate looks. Steel's browser is open source and can be self-hosted, which removes the platform fee but not infrastructure cost.
Browserless (OFFICIAL, browserless.io, current published rate card). Units of 30 seconds of connection time, so one browser-hour equals 120 units. Scale: $350 a month on annual billing, 500,000 units included, then $0.0015 per unit overage. Browserless's documentation states that proxy traffic also consumes units under the platform's official unit-consumption rules, but does not publish the specific conversion (for example, units per MB or GB of proxy bandwidth) precisely enough to compute a confident dollar figure for proxy-heavy usage from public information alone.
Normalizing the workload
All ILLUSTRATIVE. 10,000 browser-hours a month of headless Chromium, in sessions averaging 15 minutes (40,000 sessions); average concurrency of about 14, with a peak requirement of about 40; no self-hosting. At 120 units per hour, this is 1,200,000 Browserless units. Scenario A: browser compute only, proxy within each plan's inclusion. Scenario B: 25% of browser-hours (2,500 hours) use a residential proxy at 0.5 GB per hour, or 1,250 GB of proxy traffic — a moderate figure for pages of a few megabytes loaded every 20 seconds. The peak concurrency requirement of about 40 rules out Steel's Launch tier (10 concurrent) regardless of its lower per-hour rate, so Steel Scale is used throughout.
Cost at the normalized workload
Formulas: Browserbase Startup = $99 + (10,000 − 500) × $0.10 + max(0, proxy GB − 5) × $10; Steel Scale = $250 + max(0, (10,000 × $0.08 + proxy GB × $6) − $100 monthly credit); Browserless Scale = $350 + max(0, units − 500,000) × $0.0015, browser-time portion only.
| Platform | Scenario A: browser compute, minimal proxy | Scenario B: plus 1,250 GB residential proxy |
|---|---|---|
| Steel Scale (100 concurrent, fits this workload's peak) | $950 ($250 base + $700 net usage, after the $100 monthly credit, on $800 of browser-hour usage) | $8,450 ($250 base + $8,200 net usage, after the $100 credit, on $8,300 of combined usage) |
| Browserbase Startup | $1,049 ($99 + $950 overage) | $13,499 ($1,049 + $12,450 proxy overage) |
| Browserless Scale | $1,400 ($350 base + $1,050 overage on 700,000 units beyond the 500,000 included) — browser time only, confirmed | $1,400 confirmed for browser time, plus an unknown additional unit cost for 1,250 GB of proxy traffic, since Browserless's public pricing does not detail the proxy-to-unit conversion rate |
Effective browser-hour rates in Scenario A: $0.095 on Steel Scale (all-in, including the $250 base fee and the $100 monthly credit), $0.105 on Browserbase Startup, $0.140 on Browserless Scale — all three now calculable and within a narrow band. Steel's raw usage rate is $0.08/browser-hour; the $100 monthly credit partially offsets the $250 base fee, but the all-in effective rate at this volume ($0.095/hour) still sits above the raw $0.08/hour usage rate, not below it.
Proxy bandwidth is the line item that decides the bill, where it can be priced
In Scenario B, proxy traffic is 89% of Steel's $8,450 total and 92% of Browserbase's $13,499 total. Browser time is a few cents an hour on both; residential bandwidth costs $6–$12 per GB depending on vendor and tier. The per-GB gap between Steel Scale ($6) and Browserbase Startup ($10) is large enough on its own to matter more than either vendor's hourly compute rate. Browserless cannot be placed in this comparison with confidence for the proxy-heavy scenario, specifically because its public pricing states that proxy traffic consumes units without disclosing the rate at which it does so — a real limitation of its current publicly available pricing information, not a judgment that the cost is necessarily high or low.
Concurrency and session gates
Concurrency is a plan gate, not a rate, and it can eliminate an otherwise cheaper option outright. This workload's peak of about 40 concurrent browsers fits Browserbase Startup (100) and Steel Scale (100), but not Steel Launch (10 concurrent) or Browserbase Developer (25) — meaning Steel's lower Launch-tier rate is not actually available to a workload with this concurrency profile. Browserbase Startup adds session creation limits of 50 a minute, so spinning up 100 sessions takes about two minutes. Long-running authenticated agent sessions favor hour-based billing (Steel, Browserbase), while bursts of short sessions map cleanly to Browserless's 30-second units, where a 10-second job still costs one unit.
Sensitivity
- Proxy gigabytes per browser-hour. At 0.5 GB per hour on 25% of hours, proxy bandwidth is 89–92% of the bill on Steel and Browserbase; at zero it vanishes.
- Which Steel tier concurrency actually permits. Launch's 10-concurrent ceiling rules it out for this workload regardless of its cheaper per-hour rate; a lower-concurrency workload could use Launch and pay a different (higher per-hour, no base fee) structure.
- Session length. Browserless's 30-second rounding and Browserbase's reported one-minute minimum penalize very short sessions; at 15 minutes the effect is under 5%, at 20 seconds it is large.
- Residential versus datacenter proxy mix, and Browserless's undisclosed proxy-unit rate. Residential bandwidth on Steel and Browserbase is priced at $6–$12 per GB; Browserless's equivalent rate is not published, which is itself a real sourcing gap for any team evaluating it for a proxy-heavy workload.
Budgeting traps
- Comparing hourly rates without checking concurrency ceilings first. Steel's cheaper Launch tier is unavailable at this workload's required concurrency; the rate that matters is Scale's.
- Comparing hourly rates and ignoring proxy rates. A $0.08 versus $0.10 hourly difference is a few hundred dollars here; a $6 versus $10 per-GB difference is thousands.
- Treating Steel's $100 monthly usage credit as a one-time signup bonus. It recurs monthly on the Scale plan and materially lowers the effective per-hour rate at this volume; Launch's $30 credit, by contrast, is a one-time signup credit only.
- Assuming Browserless's browser-time cost is unknowable. Its current published $0.0015-per-unit overage rate makes browser-time cost directly calculable; only the proxy-unit conversion remains undisclosed.
- Forgetting call-based meters on Browserbase. Search, Fetch and Extract calls are billed separately from browser hours.
What to ask before you buy
Ask Browserless directly for its official unit-consumption rule for proxy traffic, since the browser-time portion of its pricing is now fully calculable but the proxy portion is not disclosed publicly. Confirm Steel's current concurrency ceiling on each tier against your actual peak requirement before comparing per-hour rates, since a cheaper tier that can't support your concurrency isn't actually available to you. Measure your real proxy gigabytes per hour and session-length distribution before comparing any of these bills.