CHOOSE YOUR AI SETUP
Compare what providers publish, what Kingy measured and what readers experienced. Choose a starting point for your own task.
US personal web subscriptions. Prices exclude taxes where the source does not include them. A plan gives access to features; it does not guarantee completed tasks. Original evidence dates and unresolved allowances are retained.
USD 0.00 / free
Last verified: 2026-10-06. Evidence: source-verified.
Separate tool limits apply; specific quotas are not assumed.
Feature access is evidenced; no guaranteed completed-task capacity has been established.
Price sources: ChatGPT Free Tier FAQ
USD 20.00 / monthly
Last verified: 2026-10-06. Evidence: source-verified.
Limits and model access may vary by account and system demand.
API use is billed separately. Annual billing is unavailable.
Authorized participating-tool use can share the applicable OpenAI allowance; it does not include separately billed generic API usage or create an extra allowance.
Feature access is evidenced; no guaranteed completed-task capacity has been established.
Price sources: ChatGPT Plus: features, monthly price and separate API billing OpenAI US plan price context
USD 100.00 / monthly
Last verified: 2026-10-07. Evidence: source-verified.
Work/Codex: no current five-hour limit; weekly account allowance remains. Chat limits are separate.
Prices are US web subscription prices. Credits and API billing are separate.
Astra Ultrafast access is included only in Pro US$500 among these individual tiers.
Feature access is evidenced; no guaranteed completed-task capacity has been established.
Price sources: OpenAI current Pro tiers, Codex access and shared usage
USD 200.00 / monthly
Last verified: 2026-10-07. Evidence: source-verified.
Work/Codex: no current five-hour limit; weekly account allowance remains. Chat limits are separate.
Prices are US web subscription prices. Credits and API billing are separate.
Astra Ultrafast access is included only in Pro US$500 among these individual tiers.
Feature access is evidenced; no guaranteed completed-task capacity has been established.
Price sources: OpenAI current Pro tiers, Codex access and shared usage
USD 500.00 / monthly
Last verified: 2026-10-07. Evidence: source-verified.
Work/Codex: no current five-hour limit; weekly account allowance remains. Chat limits are separate.
Prices are US web subscription prices. Credits and API billing are separate.
Astra Ultrafast access is included only in Pro US$500 among these individual tiers.
Feature access is evidenced; no guaranteed completed-task capacity has been established.
Price sources: OpenAI current Pro tiers, Codex access and shared usage
Undisclosed
Current public documentation does not establish one fixed ChatGPT Chat allowance for this plan/model surface. Work and Codex estimates must not be treated as a Chat cap; use the account usage notice.
Last verified: 2026-08-02. Evidence tier: undisclosed. Surface: Assistant.
Unverified since 2026-10-08.
Reset rule: plan and model specific. Reset type: none.
When reached: The product shows the applicable reset when reached.
Undisclosed
Current public documentation does not establish one fixed ChatGPT Chat allowance for this plan/model surface. Work and Codex estimates must not be treated as a Chat cap; use the account usage notice.
Last verified: 2026-08-02. Evidence tier: undisclosed. Surface: Assistant.
Unverified since 2026-10-08.
Reset rule: plan and model specific. Reset type: none.
When reached: The product shows the applicable reset when reached.
Up to 80 files per 3 hours
Official FAQ: up to 80 files every 3 hours; limits may be lower during peak hours. This is a rolling upload-rate cap, separate from file-size and user/org storage caps. Failed uploads can count; remaining rolling quota is not displayed.
Last verified: 2026-10-06. Evidence tier: documented. Surface: Assistant.
Reset rule: rolling 3-hour upload window; storage caps also apply. Reset type: none.
When reached: Wait for the upload window or clear shared storage as applicable; failed attempts can count.
Plan and model specific
Allowance depends on the plan, model, workload and speed mode. Work/Codex usage estimates are not fixed ChatGPT Chat message limits; check the correct account surface.
Last verified: 2026-10-01. Evidence tier: undisclosed. Surface: Assistant.
Unverified since 2026-10-08.
Reset rule: varies. Reset type: none.
When reached: Model and abuse guardrails may apply.
USD 0.00 / free
Last verified: 2026-10-06. Evidence: source-verified.
Limited usage. Workload capacity is not guaranteed.
Feature access is evidenced; no guaranteed completed-task capacity has been established.
Price sources: Claude pricing
USD 20.00 / monthly
Last verified: 2026-10-06. Evidence: source-verified.
Usage limits apply; more usage does not establish a fixed task quota.
API billing and optional extra usage are separate.
Feature access is evidenced; no guaranteed completed-task capacity has been established.
Price sources: Claude pricing Claude Pro US pricing and usage limits
USD 200.00 / annual
Last verified: 2026-10-06. Evidence: source-verified.
Usage limits apply; more usage does not establish a fixed task quota.
API billing and optional extra usage are separate.
Feature access is evidenced; no guaranteed completed-task capacity has been established.
Price sources: Claude pricing Claude Pro US pricing and usage limits
USD 100.00 / monthly
Last verified: 2026-10-07. Evidence: source-verified.
5× Pro allowance per session, plus weekly and possible model/feature controls.
Claude Code access is included; mobile prices can differ. Eligible accounts can receive conditional monthly API credits, but Console/API access and billing remain separate after those credits are used.
Session allowance multiplier is documented; no guaranteed completed-task capacity established.
Price sources: Claude Max web pricing and Claude Code access
USD 200.00 / monthly
Last verified: 2026-10-07. Evidence: source-verified.
20× Pro allowance per session, plus weekly and possible model/feature controls.
Claude Code access is included; mobile prices can differ. Eligible accounts can receive conditional monthly API credits, but Console/API access and billing remain separate after those credits are used.
Session allowance multiplier is documented; no guaranteed completed-task capacity established.
Price sources: Claude Max web pricing and Claude Code access
Undisclosed
Message capacity varies with files, conversation length, model and feature. A five-hour session is documented; a universal message count or reset anchored to the moment of refusal is not.
Last verified: 2026-10-01. Evidence tier: undisclosed. Surface: Assistant.
Unverified since 2026-10-08.
Reset rule: 5-hour sessions; use the account reset time. Reset type: none.
When reached: Wait for the displayed reset or use separately paid usage credits if enabled.
Undisclosed
The weekly allowance resets at an account-assigned day and time. See Settings > Usage; it is not seven days after your first prompt or refusal.
Last verified: 2026-10-01. Evidence tier: undisclosed. Surface: Assistant.
Unverified since 2026-10-08.
Reset rule: fixed weekly reset assigned to your account. Reset type: none.
When reached: Wait for the account weekly reset or check separately paid usage credits.
Undisclosed
Max has a weekly all-model allowance and may have additional model, feature, weekly or monthly caps. Check the affected allowance in Settings > Usage.
Last verified: 2026-10-01. Evidence tier: undisclosed. Surface: Assistant.
Unverified since 2026-10-08.
Reset rule: vendor may apply weekly or monthly caps. Reset type: none.
When reached: Affected model or feature pauses until its displayed reset.
5× Pro per-session usage
A multiplier is not a guaranteed message or completed-task count. Model, context and workload change consumption.
Last verified: 2026-10-07. Evidence tier: documented. Surface: Assistant.
Reset rule: 5-hour session plus weekly controls. Reset type: not recorded.
When reached: Check every applicable account window and the displayed reset. Extra usage is separate..
20× Pro per-session usage
The multiplier is per session, not an exact weekly token quota. Other model and feature limits may apply. Community token equivalents are conditional workload reports.
Last verified: 2026-10-07. Evidence tier: documented. Surface: Agent.
Reset rule: 5-hour session plus weekly controls. Reset type: not recorded.
When reached: Check every applicable account window and the displayed reset. Extra usage is separate..
1,000 per minute
Standard maximum for this usage tier, not guaranteed capacity; actual organization limits can differ. Opus 5 has its own bucket; cached reads do not count toward its ITPM, and output counts actual generated tokens.
Last verified: 2026-10-06. Evidence tier: documented. Surface: API.
Reset rule: token bucket. Reset type: bucket.
When reached: 429 with retry-after; ramp traffic gradually.
2,000,000 per minute
Standard maximum for this usage tier, not guaranteed capacity; actual organization limits can differ. Opus 5 has its own bucket; cached reads do not count toward its ITPM, and output counts actual generated tokens.
Last verified: 2026-10-06. Evidence tier: documented. Surface: API.
Reset rule: token bucket. Reset type: bucket.
When reached: 429 with retry-after; ramp traffic gradually.
400,000 per minute
Standard maximum for this usage tier, not guaranteed capacity; actual organization limits can differ. Opus 5 has its own bucket; cached reads do not count toward its ITPM, and output counts actual generated tokens.
Last verified: 2026-10-06. Evidence tier: documented. Surface: API.
Reset rule: token bucket. Reset type: bucket.
When reached: 429 with retry-after; ramp traffic gradually.
5,000,000 per minute
Standard maximum for this usage tier, not guaranteed capacity; actual organization limits can differ. Opus 5 has its own bucket; cached reads do not count toward its ITPM, and output counts actual generated tokens.
Last verified: 2026-10-06. Evidence tier: documented. Surface: API.
Reset rule: token bucket. Reset type: bucket.
When reached: 429 with retry-after; ramp traffic gradually.
10,000,000 per minute
Standard maximum for this usage tier, not guaranteed capacity; actual organization limits can differ. Opus 5 has its own bucket; cached reads do not count toward its ITPM, and output counts actual generated tokens.
Last verified: 2026-10-06. Evidence tier: documented. Surface: API.
Reset rule: token bucket. Reset type: bucket.
When reached: 429 with retry-after; ramp traffic gradually.
$500 per calendar month
Calendar-month organization spend cap for this tier. At the cap the API returns 429 without retry-after and pauses until 00:00 UTC on the first day of the next month, unless a higher tier/limit is approved.
Last verified: 2026-10-06. Evidence tier: documented. Surface: API.
Reset rule: calendar month. Reset type: calendar-utc.
When reached: 429 without retry-after until 00:00 UTC on the first day of next month or a higher limit.
$1,000 per calendar month
Calendar-month organization spend cap for this tier. At the cap the API returns 429 without retry-after and pauses until 00:00 UTC on the first day of the next month, unless a higher tier/limit is approved.
Last verified: 2026-10-06. Evidence tier: documented. Surface: API.
Reset rule: calendar month. Reset type: calendar-utc.
When reached: 429 without retry-after until 00:00 UTC on the first day of next month or a higher limit.
$200,000 per calendar month
Calendar-month organization spend cap for this tier. At the cap the API returns 429 without retry-after and pauses until 00:00 UTC on the first day of the next month, unless a higher tier/limit is approved.
Last verified: 2026-10-06. Evidence tier: documented. Surface: API.
Reset rule: calendar month. Reset type: calendar-utc.
When reached: 429 without retry-after until 00:00 UTC on the first day of next month or a higher limit.
500 RPM / 500,000 TPM
Official GPT-5.5 API Standard limits for this exact model and usage tier. These are not GPT-6.1 Sol or ChatGPT message allowances; check your API account limits for actual capacity.
Last verified: 2026-10-06. Evidence tier: documented. Surface: API.
Reset rule: per minute. Reset type: bucket.
When reached: 429 until capacity replenishes.
5,000 RPM / 1,000,000 TPM
Official GPT-5.5 API Standard limits for this exact model and usage tier. These are not GPT-6.1 Sol or ChatGPT message allowances; check your API account limits for actual capacity.
Last verified: 2026-10-06. Evidence tier: documented. Surface: API.
Reset rule: per minute. Reset type: bucket.
When reached: 429 until capacity replenishes.
5,000 RPM / 2,000,000 TPM
Official GPT-5.5 API Standard limits for this exact model and usage tier. These are not GPT-6.1 Sol or ChatGPT message allowances; check your API account limits for actual capacity.
Last verified: 2026-10-06. Evidence tier: documented. Surface: API.
Reset rule: per minute. Reset type: bucket.
When reached: 429 until capacity replenishes.
10,000 RPM / 4,000,000 TPM
Official GPT-5.5 API Standard limits for this exact model and usage tier. These are not GPT-6.1 Sol or ChatGPT message allowances; check your API account limits for actual capacity.
Last verified: 2026-10-06. Evidence tier: documented. Surface: API.
Reset rule: per minute. Reset type: bucket.
When reached: 429 until capacity replenishes.
15,000 RPM / 40,000,000 TPM
Official GPT-5.5 API Standard limits for this exact model and usage tier. These are not GPT-6.1 Sol or ChatGPT message allowances; check your API account limits for actual capacity.
Last verified: 2026-10-06. Evidence tier: documented. Surface: API.
Reset rule: per minute. Reset type: bucket.
When reached: 429 until capacity replenishes.
Model and tier specific
Official GPT-5.5 API Standard limits for this exact model and usage tier. These are not GPT-6.1 Sol or ChatGPT message allowances; check your API account limits for actual capacity.
Last verified: 2026-10-06. Evidence tier: documented. Surface: API.
Reset rule: varies by model. Reset type: none.
When reached: Check the model page and account limits.
Model-specific
Per-project RPM/TPM/RPD vary by model and account tier; actual limits are available in AI Studio. RPD resets at midnight Pacific time, and documented capacity is not guaranteed.
Last verified: 2026-10-06. Evidence tier: console. Surface: API.
Reset rule: daily, midnight Pacific. Reset type: daily-pacific.
When reached: 429 until midnight Pacific.
$10 per 10 minutes
Tier 1 spend-based cap is $10 per rolling 10 minutes. Whether it applies depends on billing history/account standing; it is separate from token/request and billing-tier caps.
Last verified: 2026-10-06. Evidence tier: documented. Surface: API.
Reset rule: rolling 10 minutes. Reset type: rolling; window 0.16666666 hours.
When reached: Requests rejected until the window rolls.
$50 / $200 per 10 minutes
Tier 2 is $50 and Tier 3 is $200 per rolling 10 minutes. The former combined $200 figure was wrong for Tier 2. Applicability depends on billing history/account standing; billing-tier caps are separate.
Last verified: 2026-10-06. Evidence tier: documented. Surface: API.
Reset rule: rolling 10 minutes. Reset type: rolling; window 0.16666666 hours.
When reached: Requests rejected until the window rolls.
Typically instant
Free-to-Tier1 upgrades typically take effect instantly after billing setup. Subsequent tier upgrades take effect within 10 minutes; rare approval exceptions apply.
Last verified: 2026-10-06. Evidence tier: documented. Surface: API.
Reset rule: one-off. Reset type: none.
When reached: n/a.
2× free tier
Personal-account compute allowance, not a guaranteed number of prompts. Usage depends on prompt complexity, model, features and conversation length; five-hour and weekly limits apply.
Last verified: 2026-10-01. Evidence tier: documented. Surface: Assistant.
Unverified since 2026-10-08.
Reset rule: compute allowance refreshed every 5 hours until weekly limit. Reset type: none.
When reached: Check the app reset notice; a weekly cap may still apply.
4× free tier
Personal-account compute allowance, not a guaranteed number of prompts. Usage depends on prompt complexity, model, features and conversation length; five-hour and weekly limits apply.
Last verified: 2026-10-01. Evidence tier: documented. Surface: Assistant.
Unverified since 2026-10-08.
Reset rule: compute allowance refreshed every 5 hours until weekly limit. Reset type: none.
When reached: Check the app reset notice; a weekly cap may still apply.
5× or 20× AI Pro
Personal-account compute allowance, not a guaranteed number of prompts. Usage depends on prompt complexity, model, features and conversation length; five-hour and weekly limits apply.
Last verified: 2026-10-01. Evidence tier: documented. Surface: Assistant.
Unverified since 2026-10-08.
Reset rule: compute allowance refreshed every 5 hours until weekly limit. Reset type: none.
When reached: Check the app reset notice; a weekly cap may still apply.
Superseded
Historical fixed daily-prompt guidance remains superseded by the current compute-based policy. This October 1 review does not create a new daily message measurement.
Last verified: 2026-09-06. Evidence tier: observed. Surface: Assistant.
Unverified since 2026-10-08.
Reset rule: replaced in May 2026. Reset type: none.
When reached: n/a.
USD 0.00 / free
Last verified: 2026-10-06. Evidence: source-verified.
Code completions are suggestions, not finished coding tasks. A monthly allowance does not establish burst throughput or code quality.
Copilot CLI and agent usage have variable limits. Extra metered usage is outside this price.
Copilot Free is only for individual developers without assigned organization or enterprise Copilot access.
General completed-task capacity remains unknown; the separate code-completion feature allowance is documented.
Price sources: GitHub Copilot: current USD prices and feature allowances
USD 10.00 / monthly
Last verified: 2026-10-06. Evidence: source-verified.
Code completions are suggestions, not finished coding tasks. A monthly allowance does not establish burst throughput or code quality.
Copilot CLI and agent usage have variable limits. Extra metered usage is outside this price.
Published Pro credits include a variable flex allotment. No number of completed agent tasks is inferred from the credit amount.
General completed-task capacity remains unknown; the separate code-completion feature allowance is documented.
Price sources: GitHub Copilot: current USD prices and feature allowances
2,000 per month
GitHub documents 2,000 Free completions per month. This row does not infer the inline-completion reset timestamp from the separate AI-credit rules.
Last verified: 2026-10-06. Evidence tier: documented. Surface: Coding.
Reset rule: monthly; check account reset. Reset type: none.
When reached: Wait for the account allowance refresh or review upgrade options.
Amount undisclosed
Current GitHub documentation confirms a Free AI-credit allowance but does not publish its amount.
Last verified: 2026-10-06. Evidence tier: undisclosed. Surface: Coding.
Reset rule: calendar month. Reset type: calendar-utc.
When reached: Upgrade or wait for the monthly reset.
1,500 credits per month
The current paid allowance includes base credits and a variable flex allotment: Pro 1,500, Pro+ 7,000, Max 20,000. AI credits reset at 00:00 UTC on the first day of the month and do not roll over. Check the current flex allocation in account usage.
Last verified: 2026-10-06. Evidence tier: documented. Surface: Coding.
Reset rule: calendar month. Reset type: calendar-utc.
When reached: Buy additional credits or wait for reset.
7,000 credits per month
The current paid allowance includes base credits and a variable flex allotment: Pro 1,500, Pro+ 7,000, Max 20,000. AI credits reset at 00:00 UTC on the first day of the month and do not roll over. Check the current flex allocation in account usage.
Last verified: 2026-10-06. Evidence tier: documented. Surface: Coding.
Reset rule: calendar month. Reset type: calendar-utc.
When reached: Buy additional credits or wait for reset.
20,000 credits per month
The current paid allowance includes base credits and a variable flex allotment: Pro 1,500, Pro+ 7,000, Max 20,000. AI credits reset at 00:00 UTC on the first day of the month and do not roll over. Check the current flex allocation in account usage.
Last verified: 2026-10-06. Evidence tier: documented. Surface: Coding.
Reset rule: calendar month. Reset type: calendar-utc.
When reached: Buy additional credits or wait for reset.
300 per month
Legacy request-based plan: 300 premium requests per month; confirm your account still uses this billing model.
Last verified: 2026-10-06. Evidence tier: documented. Surface: Coding.
Reset rule: calendar month, 00:00 UTC on the 1st. Reset type: calendar-utc.
When reached: Continues on an included model.
1,500 per month
Legacy request-based plan: 1,500 premium requests per month; confirm your account still uses this billing model.
Last verified: 2026-10-06. Evidence tier: documented. Surface: Coding.
Reset rule: calendar month, 00:00 UTC on the 1st. Reset type: calendar-utc.
When reached: Continues on an included model.
13 requests per review
Under legacy request-based billing, each code review consumes 13 premium requests. AI-credit billing is a different system.
Last verified: 2026-10-06. Evidence tier: documented. Surface: Coding.
Reset rule: drawn from monthly allowance. Reset type: calendar-utc.
When reached: Counts against the monthly pool.
Undisclosed
Current plan table documents the pool inclusion, with no guaranteed numeric amount. Pools reset with the monthly billing cycle; see the editor/usage dashboard. Start (India) excludes Other Models.
Last verified: 2026-10-02. Evidence tier: undisclosed. Surface: Coding.
Unverified since 2026-10-09.
Reset rule: billing cycle. Reset type: billing.
When reached: Check the affected pool in account usage; eligible paid plans can use on-demand usage.
Undisclosed
Current plan table documents the pool inclusion, with no guaranteed numeric amount. Pools reset with the monthly billing cycle; see the editor/usage dashboard. Start (India) excludes Other Models.
Last verified: 2026-10-02. Evidence tier: undisclosed. Surface: Coding.
Unverified since 2026-10-09.
Reset rule: billing cycle. Reset type: billing.
When reached: Check the affected pool in account usage; eligible paid plans can use on-demand usage.
Undisclosed
Current plan table documents the pool inclusion, with no guaranteed numeric amount. Pools reset with the monthly billing cycle; see the editor/usage dashboard. Start (India) excludes Other Models.
Last verified: 2026-10-02. Evidence tier: undisclosed. Surface: Coding.
Unverified since 2026-10-09.
Reset rule: billing cycle. Reset type: billing.
When reached: Check the affected pool in account usage; eligible paid plans can use on-demand usage.
Undisclosed
Current plan table documents the pool inclusion, with no guaranteed numeric amount. Pools reset with the monthly billing cycle; see the editor/usage dashboard. Start (India) excludes Other Models.
Last verified: 2026-10-02. Evidence tier: undisclosed. Surface: Coding.
Unverified since 2026-10-09.
Reset rule: billing cycle. Reset type: billing.
When reached: Check the affected pool in account usage; eligible paid plans can use on-demand usage.
50 per day
Official documentation specifies 50 or 1,000 free-model requests per UTC day, with a nominal 10-credit purchase threshold. The higher daily ceiling starts at 9 purchased credits to absorb fees and rounding. Check the actual daily counter and ceiling with GET /api/v1/key; credit balance, per-key caps and provider limits are separate.
Last verified: 2026-10-06. Evidence tier: documented. Surface: API.
Reset rule: daily. Reset type: daily-utc.
When reached: 429 with X-RateLimit headers.
1,000 per day
Official documentation specifies 50 or 1,000 free-model requests per UTC day, with a nominal 10-credit purchase threshold. The higher daily ceiling starts at 9 purchased credits to absorb fees and rounding. Check the actual daily counter and ceiling with GET /api/v1/key; credit balance, per-key caps and provider limits are separate.
Last verified: 2026-10-06. Evidence tier: documented. Surface: API.
Reset rule: daily. Reset type: daily-utc.
When reached: 429 with X-RateLimit headers.
20 per minute
Official documentation specifies 20 free-model requests per minute. This platform free-variant cap is separate from paid-provider limits, credit balance and DDoS protection.
Last verified: 2026-10-06. Evidence tier: documented. Surface: API.
Reset rule: per minute. Reset type: bucket.
When reached: 429 with X-RateLimit headers.
00:00 UTC
The official GET /api/v1/key schema documents free_model_daily_requests used, limit and remaining for the current UTC day. This establishes the daily counter boundary; per-key spending-limit resets are separate.
Last verified: 2026-10-06. Evidence tier: documented. Surface: API.
Reset rule: UTC calendar day. Reset type: daily-utc.
When reached: n/a.
500 concurrent connections
Account-level concurrency: deepseek-v4-pro 500; deepseek-flash 2,500. A slot is occupied from send until the response completes. All API keys share the account limit.
Last verified: 2026-10-06. Evidence tier: documented. Surface: API.
Reset rule: concurrent, not time-windowed. Reset type: concurrency.
When reached: HTTP 429.
2,500 concurrent connections
Account-level concurrency: deepseek-v4-pro 500; deepseek-flash 2,500. A slot is occupied from send until the response completes. All API keys share the account limit.
Last verified: 2026-10-06. Evidence tier: documented. Surface: API.
Reset rule: concurrent, not time-windowed. Reset type: concurrency.
When reached: HTTP 429.
Not published
The current public reference specifies concurrency instead of a universal RPM/TPM quota; no numeric RPM/TPM cap is stated.
Last verified: 2026-10-06. Evidence tier: undisclosed. Surface: API.
Reset rule: n/a. Reset type: none.
When reached: n/a.
500 / 2,500 per user_id
For expanded-quota accounts, each user_id has 500 deepseek-v4-pro / 2,500 deepseek-flash concurrent connections while the total account concurrency limit also applies. For regular users, user_id values share the account pool.
Last verified: 2026-10-06. Evidence tier: documented. Surface: API.
Reset rule: concurrent, not time-windowed. Reset type: concurrency.
When reached: HTTP 429 for that user_id only.
150 RPS / 50M TPM
Published Tier 0 grok-4.6 maximum: 150 requests/second and 50 million tokens/minute. The per-second limit is derived from the per-minute request budget (RPM / 48, with a minimum of 2). Limits are per model and scale with team tier; check the xAI Console for personalized limits.
Last verified: 2026-10-01. Evidence tier: documented. Surface: API.
Unverified since 2026-10-08.
Reset rule: per second / per minute. Reset type: bucket.
When reached: 429 until capacity replenishes.
$20 per month
The current Core pricing card includes $20 toward paid models. This differs from the former $25 allocation. Current pricing is $20 monthly or $18/month billed annually; tax can apply.
Last verified: 2026-10-01. Evidence tier: documented. Surface: Coding.
Unverified since 2026-10-08.
Reset rule: monthly. Reset type: billing.
When reached: Paid work draws on credits; balance and configured budgets can stop work.
$100 per month
The current base Pro pricing card includes $100 toward paid models; the Pro billing guide describes multiple credit tiers. This is the base-tier amount, not a guarantee for every Pro subscription.
Last verified: 2026-10-01. Evidence tier: documented. Surface: Coding.
Unverified since 2026-10-08.
Reset rule: monthly. Reset type: billing.
When reached: Paid work draws on credits; balance and configured budgets can stop work.
Undisclosed
The current Replit pricing page lists Core, Pro and Enterprise; it does not list Starter. No current Starter daily Agent allowance is established by this source. Do not transfer a former Starter allowance to Core.
Last verified: 2026-10-01. Evidence tier: undisclosed. Surface: Coding.
Unverified since 2026-10-08.
Reset rule: daily. Reset type: none.
When reached: Agent unavailable until the next day.
Two months
The current Pro billing guide states that unused credits roll over for two months. This corrects the former one-month statement.
Last verified: 2026-10-02. Evidence tier: documented. Surface: Coding.
Unverified since 2026-10-09.
Reset rule: billing cycle. Reset type: billing.
When reached: Older unused credits expire after the stated rollover.
Daily and weekly allowances
Pro and Teams full seats have daily and weekly allowances; Max has a weekly allowance without a daily cap. All refresh on the documented calendar basis. The allowance is shared across sessions, CLI and Desktop; no fixed public message count.
Last verified: 2026-10-02. Evidence tier: documented. Surface: Coding.
Unverified since 2026-10-09.
Reset rule: calendar cadence. Reset type: none.
When reached: Wait for the account reset or use separately purchased on-demand credits if enabled.
Not published
The public billing reference does not publish a guaranteed per-message consumption amount. Included quotas and on-demand usage differ; legacy ACUs are not a new fixed message entitlement.
Last verified: 2026-10-02. Evidence tier: undisclosed. Surface: Coding.
Unverified since 2026-10-09.
Reset rule: per request. Reset type: none.
When reached: Draws down the unpublished allowance.
$20 per month
Official self-serve plan table: Pro $20/month and Max $200/month for a single user. Teams has separate seat/minimum pricing; current legacy Desktop plans can differ.
Last verified: 2026-10-02. Evidence tier: documented. Surface: Coding.
Unverified since 2026-10-09.
Reset rule: monthly. Reset type: billing.
When reached: Allowance refreshes daily and weekly; figure unpublished.
$200 per month
Official self-serve plan table: Pro $20/month and Max $200/month for a single user. Teams has separate seat/minimum pricing; current legacy Desktop plans can differ.
Last verified: 2026-10-02. Evidence tier: documented. Surface: Coding.
Unverified since 2026-10-09.
Reset rule: monthly. Reset type: billing.
When reached: Allowance refreshes daily and weekly; figure unpublished.
Renamed
The official Desktop FAQ identifies Devin Desktop as the new name for Windsurf. The IDE remains accessible; existing plan pricing, including legacy Enterprise plans, is preserved.
Last verified: 2026-10-06. Evidence tier: documented. Surface: Coding.
Reset rule: n/a. Reset type: none.
When reached: n/a.
50 per minute
Historical Sonar quota observed in August; it is not a current Agent API entitlement. Sonar support ended September 27. Synchronous and streaming compatibility requests are being reformulated by model; asynchronous requests are no longer supported. No new inference call was made.
Last verified: 2026-08-27. Evidence tier: documented. Surface: API.
Unverified since 2026-10-08.
Reset rule: per minute. Reset type: none.
When reached: Historical quota; use current Agent API/account limits.
5 per minute
Historical Sonar quota observed in August; it is not a current Agent API entitlement. Sonar support ended September 27. Synchronous and streaming compatibility requests are being reformulated by model; asynchronous requests are no longer supported. No new inference call was made.
Last verified: 2026-08-27. Evidence tier: documented. Surface: API.
Unverified since 2026-10-08.
Reset rule: per minute. Reset type: none.
When reached: Historical quota; use current Agent API/account limits.
1,000 per minute
Historical Sonar quota observed in August; it is not a current Agent API entitlement. Sonar support ended September 27. Synchronous and streaming compatibility requests are being reformulated by model; asynchronous requests are no longer supported. No new inference call was made.
Last verified: 2026-08-27. Evidence tier: documented. Surface: API.
Unverified since 2026-10-08.
Reset rule: per minute. Reset type: none.
When reached: Historical quota; use current Agent API/account limits.
4,000 per minute
Historical Sonar quota observed in August; it is not a current Agent API entitlement. Sonar support ended September 27. Synchronous and streaming compatibility requests are being reformulated by model; asynchronous requests are no longer supported. No new inference call was made.
Last verified: 2026-08-27. Evidence tier: documented. Surface: API.
Unverified since 2026-10-08.
Reset rule: per minute. Reset type: none.
When reached: Historical quota; use current Agent API/account limits.
100 per minute
Historical Sonar quota observed in August; it is not a current Agent API entitlement. Sonar support ended September 27. Synchronous and streaming compatibility requests are being reformulated by model; asynchronous requests are no longer supported. No new inference call was made.
Last verified: 2026-08-27. Evidence tier: documented. Surface: API.
Unverified since 2026-10-08.
Reset rule: per minute. Reset type: none.
When reached: Historical quota; use current Agent API/account limits.
Cumulative spend
Current Perplexity API tiers use cumulative purchased credits across the account lifetime, not balance. Thresholds are $0 / $50 / $250 / $500 / $1,000 / $5,000. Reached tiers do not downgrade.
Last verified: 2026-10-02. Evidence tier: documented. Surface: API.
Unverified since 2026-10-09.
Reset rule: account lifetime. Reset type: none.
When reached: n/a.
Support ended; compatibility migration
Sonar Chat Completions support ended September 27, 2026. Synchronous and streaming requests continue through a gradual by-model reformulation as Agent API requests. Asynchronous Sonar requests are no longer supported; migrate that work to background mode. This is a documentation review, not an inference test.
Last verified: 2026-10-02. Evidence tier: documented. Surface: API.
Unverified since 2026-10-09.
Reset rule: n/a. Reset type: none.
When reached: n/a.
None no monthly allocation
Paid Consumer Pro includes weekly Computer usage, separate from credits. Without credits, included usage runs at Light effort. No monthly credit allocation; promotional or non-paying Pro is excluded. Check account usage for the amount and weekly reset.
Last verified: 2026-10-06. Evidence tier: documented. Surface: Assistant.
Reset rule: n/a. Reset type: none.
When reached: Paid Pro: use included weekly usage, then credits; without credits and auto-refill off, wait for the weekly reset or purchase credits.
4,000 one-time
The listed signup bonus is once per account and expires after 30 days; eligibility and promotions may vary.
Last verified: 2026-10-06. Evidence tier: documented. Surface: Assistant.
Reset rule: expires 30 days after signup. Reset type: none.
When reached: Not replenished.
10,000 per month
Max includes 10,000 monthly credits. Monthly credits refresh on the billing cycle date and do not carry over.
Last verified: 2026-10-06. Evidence tier: documented. Surface: Assistant.
Reset rule: monthly refresh. Reset type: billing.
When reached: Buy more, or wait for the monthly refresh.
500 per month
Enterprise Pro lists 500 monthly Computer credits; check the exact allocation in your account.
Last verified: 2026-10-06. Evidence tier: documented. Surface: Assistant.
Reset rule: monthly refresh. Reset type: billing.
When reached: Buy more, or wait for the monthly refresh.
15,000 per month
Enterprise Max lists 15,000 monthly Computer credits; check the exact allocation in your account.
Last verified: 2026-10-06. Evidence tier: documented. Surface: Assistant.
Reset rule: monthly refresh. Reset type: billing.
When reached: Buy more, or wait for the monthly refresh.
100 credits = $1
100 credits represents $1. Task cost is variable. The article contains inconsistent light-task examples, so this reference does not forecast task costs from them.
Last verified: 2026-10-06. Evidence tier: documented. Surface: Assistant.
Reset rule: n/a. Reset type: none.
When reached: n/a.
Free no credit cost
Ask searches do not use Computer credits. Search limits and Computer credit balances are separate.
Last verified: 2026-10-06. Evidence tier: documented. Surface: Assistant.
Reset rule: n/a. Reset type: none.
When reached: n/a.
3 Pro Searches per day
The current comparison table publishes 3 Pro Searches per day for Free. It does not state the reset timezone or exact timestamp.
Last verified: 2026-10-06. Evidence tier: documented. Surface: Assistant.
Reset rule: daily; boundary not published here. Reset type: none.
When reached: Use basic search or check the account reset.
Amount undisclosed
Consumer Pro/Max Pro Search is described using weekly average/advanced-use limits; no universal numeric weekly consumer cap is stated. Enterprise limits are separate.
Last verified: 2026-10-06. Evidence tier: undisclosed. Surface: Assistant.
Reset rule: weekly; account-specific. Reset type: none.
When reached: Unknown until encountered.
3.3 hours per month
Published monthly Fast GPU allowance. This measures compute time, not finished images or elapsed waiting time. Check Manage Subscription for your renewal and balance.
Last verified: 2026-10-06. Evidence tier: documented. Surface: Creative.
Reset rule: monthly, does not carry over. Reset type: billing.
When reached: Use eligible Relax mode if your plan supports the task, buy extra Fast time, or wait for renewal.
15 hours per month
Published monthly Fast GPU allowance. This measures compute time, not finished images or elapsed waiting time. Check Manage Subscription for your renewal and balance.
Last verified: 2026-10-06. Evidence tier: documented. Surface: Creative.
Reset rule: monthly, does not carry over. Reset type: billing.
When reached: Use eligible Relax mode if your plan supports the task, buy extra Fast time, or wait for renewal.
30 hours per month
Published monthly Fast GPU allowance. This measures compute time, not finished images or elapsed waiting time. Check Manage Subscription for your renewal and balance.
Last verified: 2026-10-06. Evidence tier: documented. Surface: Creative.
Reset rule: monthly, does not carry over. Reset type: billing.
When reached: Use eligible Relax mode if your plan supports the task, buy extra Fast time, or wait for renewal.
60 hours per month
Published monthly Fast GPU allowance. This measures compute time, not finished images or elapsed waiting time. Check Manage Subscription for your renewal and balance.
Last verified: 2026-10-06. Evidence tier: documented. Surface: Creative.
Reset rule: monthly, does not carry over. Reset type: billing.
When reached: Use eligible Relax mode if your plan supports the task, buy extra Fast time, or wait for renewal.
Unlimited
Standard, Pro and Mega include unlimited Relax image generations. Unlimited Relax video is a different entitlement restricted to Pro/Mega and SD video.
Last verified: 2026-10-06. Evidence tier: documented. Surface: Creative.
Reset rule: queued, not metered. Reset type: none.
When reached: Never exhausts; jobs queue instead.
$4 per hour
Purchased Fast hours expire after 60 days and require an active subscription. The plan comparison lists $4 per hour; this is separate from monthly included time.
Last verified: 2026-10-06. Evidence tier: documented. Surface: Creative.
Reset rule: purchased hours expire after 60 days. Reset type: none.
When reached: Buy more.
2× Fast time per image
Turbo consumes twice the Fast GPU time per image where supported. The current page explicitly excludes V8.1; verify model support before planning around Turbo.
Last verified: 2026-10-06. Evidence tier: documented. Surface: Creative.
Reset rule: drawn from monthly Fast allowance. Reset type: billing.
When reached: Draws the Fast allowance down twice as quickly.
3 Fast / plan-dependent Relax concurrent
Basic permits 3 Fast image jobs; Standard permits 3 Fast or Relax image jobs. Basic has no Relax entitlement; video concurrency is separate.
Last verified: 2026-10-06. Evidence tier: documented. Surface: Creative.
Reset rule: concurrent, not time-windowed. Reset type: concurrency.
When reached: Further prompts queue.
12 Fast / 3 Relax concurrent (Fast)
Pro and Mega allow 12 concurrent Fast image jobs or 3 Relax image jobs. Video concurrency has a separate table; do not apply the 12-image figure to Relax or video.
Last verified: 2026-10-06. Evidence tier: documented. Surface: Creative.
Reset rule: concurrent, not time-windowed. Reset type: concurrency.
When reached: Further prompts queue.
1 hour per day
The top 2,000 image raters each day receive one extra Fast hour. Awarded time requires an active subscription and expires 30 days after award.
Last verified: 2026-10-06. Evidence tier: documented. Surface: Creative.
Reset rule: daily, top 2,000 raters. Reset type: none.
When reached: n/a.
5–45 local Astra messages (estimate)
Message estimates vary with context, tools, model and effort; the upper endpoint is not guaranteed. Work and Codex share capacity; Chat is separate.
Last verified: 2026-10-07. Evidence tier: documented. Surface: Agent.
Reset rule: 5-hour plus weekly controls. Reset type: not recorded.
When reached: Check every applicable account window and the displayed reset. Extra usage is separate..
Account-specific
Pro has no current five-hour Work/Codex limit. Weekly allowance still applies. The historical Kingy test did not identify the exact Pro price tier.
Last verified: 2026-10-07. Evidence tier: documented. Surface: Agent.
Reset rule: weekly; no current 5-hour limit. Reset type: not recorded.
When reached: Check every applicable account window and the displayed reset. Extra usage is separate..
USD 0.00 / free
Last verified: 2026-10-06. Evidence: source-verified.
Compute-based standard limits; some features can be unavailable during high demand.
Feature access is evidenced; no guaranteed completed-task capacity has been established.
Price sources: Gemini Apps limits and upgrades
USD 19.99 / monthly
Last verified: 2026-10-06. Evidence: source-verified.
Compute-based limits vary by task, model and conversation.
Extra AI credits are excluded. Annual billing exists, but this catalog has no verified annual price.
The reviewed USD 19.99 renewal quote is for the base 5 TB Google AI Pro offer; other storage bundles and promotions are outside this catalog.
Feature access is evidenced; no guaranteed completed-task capacity has been established.
Price sources: Google official US AI Pro price context
2× / 4× / 5× or 20× Pro
Plus is 2× standard, Pro 4× standard. Ultra is 5× or 20× Pro depending on the subscription. Complexity and features affect consumption; no guaranteed prompts per day.
Last verified: 2026-10-07. Evidence tier: documented. Surface: Assistant.
Reset rule: 5-hour and weekly compute pools. Reset type: not recorded.
When reached: Check every applicable account window and the displayed reset. Extra usage is separate..
32K / 128K / 1M / 1M tokens
Context capacity describes what fits in a conversation; it does not reset usage allowance or establish retrieval quality. Personal-account scope; workplace plans differ.
Last verified: 2026-10-07. Evidence tier: documented. Surface: Assistant.
Reset rule: per conversation. Reset type: not recorded.
When reached: Check every applicable account window and the displayed reset. Extra usage is separate..
This catalog renders the existing dated source records. Limits vary by product, model, account and workload. Unknown values remain unknown. Capture dates describe retained documents, not fresh account tests. The interactive workspace provides filters, comparisons, the quiz and separately labelled Kingy observations and reader reports.
Catalog release: 2026-10-11.claude-api-credits-v1. Download the original source catalog.