Recommended Free Tools
iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
Lower BigQuery AI spending by shrinking what reaches the model and by controlling how many rows each job submits. The practical sequence is: filter and materialize eligible rows in ordinary SQL, project only the fields needed for the task, remove repeated prompt text, batch homogeneous work, and enforce daily input/output-token quotas. Then measure the two bills separately—BigQuery bytes processed and remote model inference—because reducing one does not automatically reduce the other.
This guide covers AI.GENERATE and AI.GENERATE_TEXT, including when to use each, how to avoid accidental full-table scans, and how to resume safely after quota exhaustion.
Understand the two billing layers before changing SQL
A BigQuery AI query can create costs in two independent systems:
- BigQuery analysis charges: BigQuery bills the bytes read by the SQL plan. A query that scans a large source table can be expensive even if the model receives only a few rows.
- Remote model charges: The model endpoint bills inference based on the data sent for generation. Prompt text, row content and generated output consume tokens. Google Cloud documentation says calls using Gemini 2.0 or later use batch API pricing in the compared functions; verify the current model and region pricing before production use.
Prompt trimming mainly reduces model input tokens. A tighter WHERE clause, a narrower projection and a materialized staging table can reduce both the rows sent to the model and, when the source scan is also narrowed, BigQuery bytes processed. Always inspect the BigQuery job’s bytes-processed estimate or final billing record separately from remote endpoint usage.
#1 Best Overall
- Powerful Turbo Fan:CAGIWIRU Electric Air Duster have reached speed up to 130000RPM,Can efficient remove dust and debris.With LED light and Three Adjustable Speed to meet with different cleaning tasks.
- With Attached USB C Fasting Charging cable.Charing 3 hours,Enjoy up to 240 minutes of use on the lowest setting, For third speed can use for 25mins.with four charging options to meet with your needs.
- Wide Application: attached 5 different nozzles,making it suitable for a variety of scenes, such as car,pet, pc, keyboards, and other electronic devices. It also serves for office and home clean duster.
- Handy Design & Portable:Cordless air duster electric nozzle is Light weight and small size, design and comfortable to hold and space saving, convenient to carry around. You can put it in your room or inside the car without taking up space.
- Economical &Reusable: Built durable materials with a long-lasting battery,our CAGIWIRU air duster it's a sustainable alternative to disposable air cans.
Trim the prompt without removing information the model needs
Model input is billable. Token quotas count the instruction and the other data supplied to the model, so every repeated sentence, unused column and oversized example has a cost. Build the smallest prompt that still makes the output reliable.
Keep one concise instruction
State the task, the allowed output and the success conditions once. For a classification task, an instruction such as Classify this record in one label. Return JSON. is usually more economical than repeating formatting rules for every field. Put stable guidance in the instruction and variable content in clearly delimited fields.
Send only required columns
Do not pass an entire record when the decision depends on two or three attributes. Select those attributes in the staging query and construct the prompt there. Excluding internal IDs, audit columns, long descriptions, unused metadata and duplicated text cuts input tokens before the remote call.
Control examples and visual payloads
Few-shot examples can improve consistency, but they should represent the edge cases that matter rather than repeat the same pattern. Remove examples that do not change the decision. Google Cloud guidance gives a limit of fewer than 16,000 tokens for a visual-content prompt; treat that as a documented constraint for that workload, not as a general token budget for every model.
Make the output contract explicit
Specify a compact format and allowed values. If the function and model support a schema-constrained response, a schema can avoid verbose prose and make downstream storage predictable. Do not ask for explanations, citations or multiple representations when the pipeline only needs a label or a short field.
Filter rows in SQL before invoking an AI function
Eligibility logic belongs in ordinary SQL, before the AI function. Use predicates that can be evaluated without model calls, remove duplicates, and project only the fields needed for the prompt. Typical filters include a status flag, a processing watermark, non-null text and a partition or date range.
Rank #2
- Screen Cleaner is Your Screen’s New Bestfriend! - EVEO is proud to present a screen cleaner that’s perfectly suited for your TV. Meet your new screen cleaner, computer screen cleaner, laptop cleaner, iPad cleaner, Macbook,tv cleaner or any other electronic device. Our screen cleaner spray and wipe along with the included plush microfiber cloth, easily removes dust, fingerprints & other smudges on the screen’s surface. Clean effectively for a streak-free screen
- Gentle on your Screens - Our screen cleaner spray and wipes is compatible to be used even for the most sensitive LCD, LED, CRT, and OLED screens in proper use. Be at ease knowing that this product is compatible with all the major brands: LG, SONY, Samsung, Sharp, iPad, iPhone, Android screens, Kindle, PC monitors - we’ve got you covered! Can be used as LCD screen cleaner, tv cleaner, smart tv screen cleaner, and more
- Premium Super Soft Plush Microfiber Cloth - Included in this efficient screen cleaning kit. This screen cleaner spray and wipe has our signature plush microfiber cloth specially designed to comply with screens and monitors and leave them streak-free. Has been shown to effectively absorbs stubborn stains, zero streaks, and no fiber shedding. Always use this together with our screen cleaner for the best results
- Easy-to-Use Screen Cleaner Kit + microfiber cleaning cloth - The EVEO Screen Cleaning Kit will easily be your new favorite tool when it comes to taking care of the things you love most, quick and easy solution for a streak-free screen. Simply screw on the sprayer gun on the screen cleaner bottle. Spray the cleaning liquid onto the microfiber cloth, then wipe the screen until desired cleanliness & shine is achieved.
- Quality Screen Care Kit Worth Purchasing - Enjoy a convenient Screen Cleaner anytime, anywhere you need it. Your satisfaction is our top priority. We stand behind the quality of our products and that’s why, you shouldn’t be worrying at all. We manufacture the Best Screen Cleaner. and can be used for laptop cleaning kit, laptop screen cleaner, tv screen cleaner for smart tv. MacBook screen cleaner, monitor cleaner or screen cleaner spray for electronic devices with microfiber cleaning cloth
Why LIMIT and OFFSET are unsafe cost controls
LIMIT and OFFSET describe the result returned by a query; they do not guarantee that BigQuery will avoid reading all qualifying input first. In an AI query, the engine may process many rows before applying the limit, so model and scan costs can exceed the apparent batch size. Use selective WHERE predicates, deduplication and a staged table instead.
Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallMaterialize an exact work queue
Write the filtered result to a temporary or permanent table, inspect its row count, and pass that table to the AI function. Materialization gives you a stable retry boundary: completed rows can be recorded, and a later run can select only rows still missing a result. Google Cloud specifically recommends writing query results to a table and referencing that table so as few rows as possible are sent to the model.
Illustrative staging query
CREATE OR REPLACE TABLE `project.dataset.ai_batch` AS
SELECT
id,
CONCAT(
'Classify this record in one label. Return JSON. Text: ',
text_column
) AS prompt
FROM `project.dataset.source`
WHERE needs_ai = TRUE
AND text_column IS NOT NULL;After creation, check the table’s row count and sample prompts. If the count is unexpectedly high, fix the eligibility query before making any remote call. Keep the original identifier so each response can be joined back to the source and marked complete.
Batch rows that share the same task
Batching is a set-based way to apply one instruction to many eligible rows. It reduces orchestration overhead and gives quotas, job time and retries a clear boundary. Homogeneous rows—such as one classification task with the same output schema—are the best candidates.
Use documented scale as a planning range
The function comparison describes processing between 1 million and 10 million rows per six-hour query job. That is a documented scale range, not a throughput promise. Actual completion time depends heavily on average input-token length, model behavior, region and concurrent service load. A batch of short prompts can complete very differently from the same row count containing long documents.
The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Choose a batch boundary deliberately
Set a boundary that fits your daily token quota, expected job duration and retry tolerance. Smaller batches limit the amount rerun after a transient failure and make quota exhaustion easier to diagnose. Larger batches reduce scheduling overhead but can leave more unfinished work when a job stops. Use a deterministic key range, date partition or hash bucket so each retry selects the same rows.
Rank #3
- 【Large size】：Measuring 800 x 300 mm/ 31. 5×12 inches, which is wider and taller than others. Large mouse pad can perfectly fit your keyboard and mouse of any size, and there is also ample space for peripherals such as phones, tablets, etc.
- 【LED backlight mouse pad】：Blade Hawks RGB mouse pad provides 7 static modes and 7 dynamic modes will give you the options you are looking for. Power-off memory function remembers the last lighting mode you selected. The super glow fiber Chroma will give you a colorful gaming setup that you want.
- 【Anti-slip rubber base】：Made from Natural rubber, this gaming mousepad will firmly grip your desktop, allowing you to enjoy the ultimate precision in the games you are most passionate about. It also comes with a texture that enhances this ability even further.
- 【Smooth and waterproof surface】: Micro-textured cloth surface, for extremely precise mouse control and a smooth movement. When liquid splashes on the gaming mouse pad, it forms water droplets and slides down. Will not delay your work or gameplay.
- 【Plug and play】: This mouse and keyboard pad is powered by USB, plug and play, no driver required. One button control light modes. Press one time to change the lighting mode, press twice to change the brightness, press and hold about 3 seconds turn ON/OFF.
Persist progress for resumable runs
Store the response, status, model identifier and processing timestamp alongside the source key. The next staging query should exclude keys with a successful result. If a job fails after partial completion, rerun only rows without a persisted response rather than submitting the entire source again.
Set token quotas as a spending boundary
Configure daily input-token and output-token quotas for the model or endpoint used by the functions. A quota is a guardrail, not a performance setting: once the daily allowance is exhausted, rows that still require LLM calls can fail with quota-exhaustion errors.
Estimate before launching
Estimate input tokens from the instruction plus the selected columns and estimate output tokens from the requested format. Multiply by the number of staged rows, then leave headroom for retries and unusually long records. If the estimate approaches the daily limit, split the work queue across days or reduce prompt and output size.
Free tools Windows power users keep installed
One-click scans. No signup required.
Handle exhaustion without losing completed work
Write results incrementally or in job-sized commits. On quota failure, preserve successful rows, record the failed keys and stop submitting new work until the quota window resets or an approved limit is increased. A later run should select only unfinished keys. Do not blindly rerun a full table after a partial quota failure.
Choose between AI.GENERATE and AI.GENERATE_TEXT
Both functions can invoke a remote generative model from BigQuery, but their input and output shapes differ. Select based on the pipeline contract, not on a presumed price advantage.
| Decision axis | AI.GENERATE | AI.GENERATE_TEXT |
|---|---|---|
| Input form | Can use a string or STRUCT prompt expression in a set-based query. |
Table-valued function that accepts a prompt table or a query producing prompts. |
| Output shape | Supports schema-constrained output, including a STRUCT, when the selected model supports it. |
Returns generated text rows; parse or transform the text downstream when needed. |
| Best fit | Tasks that need typed fields, predictable columns or a structured response contract. | Large prompt-table workflows and text generation where a table input is natural. |
| Batch design | Use a filtered, set-based query and a deterministic work queue. | Materialize a prompt table, then pass that table directly to the function. |
| Model and region support | Depends on the model and region selected; confirm current documentation. | Depends on the model and region selected; confirm current documentation. |
| Billing considerations | BigQuery bytes processed and remote inference are separate; Gemini 2.0-or-later calls use batch API pricing according to current Google Cloud comparison guidance. | Same two billing layers; prompt-table materialization can help control how many rows reach the endpoint. |
Compare average prompt tokens, expected row count, output size, quota behavior, retry strategy and bytes scanned for your workload. A function that produces a convenient schema does not make a large source scan free, and a table-valued function does not remove token charges.
Rank #4
- 【Softer and More Comfortable】Vaydeer wrist rest has unique diamond pattern, which is the combination of softness and aesthetics. The materials of wrist rest are improved into higher quality memory foam and covered with silky smooth lycra. The computer wrist rest makes you as comfortable and cushiony as like rest your wrists on clouds.
- 【Ergonomic Wrist Saver】The wrist rests for keyboard and mouse comes with a 17.32×3.15×0.83 inch keyboard wrist pad and a 5.94×3.15×0.83 inch mouse wrist support. Based on ergonomic design, the unique concave shape is the perfect fit for your wrist joints. The wrist rest pad fits most computer keyboards and laptops, improve hand and wrist posture, release your wrist and arm stress.
- 【Non-Slip Rubber Bottom】Featuring an anti-skid silicone base on the bottom, this wrist keyboard support stays firmly in place on your desk, preventing the padding from sliding around, ensuring stable and consistent wrist support during extended computer sessions.
- 【Better Experience & Pain Relief】Our keyboard arm rest is beneficial to alleviate the soreness caused by direct contact and friction between your arm and a hard desk surface, reducing the risk of wrist fatigue or carpal tunnel. The soft texture of memory foam can evenly distribute the pressure around your wrists and provide good support with just enough give.
- 【Helpful in Multiple Scenarios】Whether you're working, studying, writing, typing, gaming, this keyboard and mouse rest combo is an essential accessory to add comfort and support to your hands and wrists. It’s also a great gift for men, women, family, friend, coworker, gamer, teacher, etc.
Run a table-input batch with AI.GENERATE_TEXT
Once the work queue is materialized, the table-input pattern keeps row selection separate from inference:
SELECT *
FROM AI.GENERATE_TEXT(
MODEL `project.dataset.gemini_model`,
TABLE `project.dataset.ai_batch`
);The example is illustrative. Adapt the model name, output columns, region and invocation options to the current BigQuery documentation. The important cost control is that ai_batch contains only eligible rows and only the prompt fields required for the task.
Validate the table before submission
- Confirm the row count is the intended batch size.
- Check that prompts do not contain duplicated instructions or accidental full-record JSON.
- Measure the longest prompt and investigate outliers that could dominate token usage.
- Verify the table and model are available in compatible regions.
- Confirm the destination schema can store the generated text or status.
Measure both bills and the work actually submitted
Cost optimization is incomplete until you can attribute changes. For each run, record the source bytes processed, staged row count, average and maximum prompt tokens, generated-token total, model and region, job duration, quota remaining and number of retries.
When prompt trimming helps only one bill
Suppose a query still scans the same unpartitioned source table but sends shorter prompts. Remote inference cost can fall while BigQuery bytes processed remain unchanged. Conversely, a narrower source predicate can reduce scan bytes while a verbose prompt keeps model charges high. Treat those as separate optimization projects and report them separately.
Use a before-and-after comparison
Run the original and optimized designs on equivalent row sets. Compare staged row count, bytes processed, input tokens, output tokens and successful results—not just elapsed time. Keep model, region and output requirements constant so the difference reflects the SQL and prompt changes.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Fix the driver behind crashes, sound loss and screen glitches3Clear out junk files and repair common Windows errorsTroubleshoot unexpectedly high cost or row counts
The staged table has more rows than expected
Inspect the WHERE predicates for null handling, broad date ranges and joins that multiply records. Check whether a supposedly unique source key is duplicated. Deduplicate before constructing prompts and store the count as a run-level metric.
Best Value
- 【No Drill Mounted】Cable management tray can be assembled in as little as 3 minutes or less-simply. Side clips easily clampe the edge of desk, pass through the holes on the sides, and secure excess cables with baskets. Desk cable management supports inward or outward installation to meet your needs in different scenarios, and provides you with great convenience when collecting and organizing wires.
- 【Rubber Pad Protects Desk】Wire organizer under desk built-in soft rubber pads on the clamp protect your desk from scratches, so you will not damage your furniture or office.
- 【Sturdy and Beautiful】Cord management under desk is made of high-quality carbon steel and holds about 15 lbs. The wire basket has 2 holes on each side of the tray to enable your cables to pass freely. Cable baskets are suitable for desk thickness use from 0.4" to 2.4".
- 【Space Saving & Keep Organized】Desk cable management tray holds all your power cords, outlet strips, USB hubs and computer transformers hidden from your feet, and the data cables will not enter the vacuum cleaner to keep your desk clean and tidy. Desk wire organizer is also suitable for use in the kitchen and outdoors.
- 【Reduce Safety Risks】Cable tray under desk keeps all plugs away from your kids, no more worry about tripping. Using 2 holes wire tray saves extra space, helps you easily cabling and wire protection. The mesh design is not easy to accumulate dust, but avoids safety accidents caused by messy wires.
A LIMIT did not reduce the bill
Remove the limit-based batching approach. Create a filtered table using a selective predicate or deterministic key range, verify its count, and invoke the function against that table. Apply limits only after the eligibility set has been narrowed and materialized.
Prompt tokens are unexpectedly large
Profile each concatenated field, including hidden JSON, repeated examples and long free-text columns. Project only needed columns, truncate or summarize content when the task allows it, and move stable instructions out of per-row data. For visual prompts, keep the documented fewer-than-16,000-token constraint in view.
Quota errors appear partway through a run
Read the quota-exhaustion status, persist successful responses and record the remaining keys. Reduce the next batch, wait for the quota window or request an approved quota change. Retrying the full input can consume the remaining allowance again.
BigQuery cost is high even after reducing model usage
Review the query plan and bytes processed. Add partition or clustering predicates where appropriate, avoid selecting unused columns and materialize a narrow staging table from the smallest practical source range. Prompt changes cannot undo an unchanged table scan.
An end-to-end operating procedure
- Define the output contract. Decide whether the result is free text or a schema-constrained structure, list allowed values and set a maximum useful response size.
- Build the smallest prompt. Keep one instruction, the required format and only the row fields that affect the decision.
- Pre-filter with SQL. Apply eligibility, date or partition predicates; remove duplicates; exclude null or already-completed rows.
- Materialize the queue. Write the result to a temporary or permanent table and inspect row count, prompt lengths and sample content.
- Set a batch boundary. Choose a deterministic key range or bucket sized for token quota, job-time and retry limits.
- Select the function. Use
AI.GENERATEfor structured output needs orAI.GENERATE_TEXTfor table-oriented text generation, after confirming model and region support. - Submit and persist. Save responses, source keys, status and timestamps so completed rows are not resubmitted.
- Monitor both charges. Record BigQuery bytes processed and remote inference usage as separate metrics.
- Retry intentionally. Rerun only failed or unfinished keys, adjusting prompt size, batch size or quota plan when the failure indicates a limit.
Cost-control checklist for production
- Prompt contains no repeated instruction or unused column.
- Eligibility uses
WHERE, partition or key predicates rather than relying onLIMIT/OFFSET. - Duplicates are removed before inference.
- Staged row count is measured and approved.
- Average and maximum prompt tokens are tracked.
- Input and output daily quotas are configured with operational headroom.
- Successful results are persisted before the next batch.
- Retries select unfinished keys only.
- BigQuery bytes and remote model charges are reviewed independently.
- Model, region, pricing and function limits are rechecked against current Google Cloud documentation before rollout.
What this approach can and cannot guarantee
Trimming prompts and batching rows reliably reduce avoidable input and orchestration overhead when the same task is applied to many records. They do not guarantee a particular completion time, throughput or price: token lengths, output variability, model availability, region and service conditions affect every run. The documented 1-million-to-10-million-row, six-hour range is useful for capacity planning, while your measured prompt distribution and quota settings determine the practical batch size.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
