Week 41

Lower cache costs. Claimable API credits.

October 5-11, 2026 UTC. Sonnet cache reads cost less, eligible Claude subscribers can claim API credits, and Gemini access changes need an account check.

One component price cutOfficial sourcesReviewed October 13

What changed for your wallet

01
Sonnet cache reads halve. Direct Claude Sonnet 5.5 cache reads fall from $0.20 to $0.10 per million tokens. At 100M cache-read tokens, that component costs $10 instead of $20.
02
Claim eligible credits before buying API balance. Max 5x/20x includes $100/$200 in monthly direct API credits after eligibility and linking. This can reduce separately purchased usage, without changing subscription limits.
03
Check Gemini access before upgrading. Free Auto can still route harder prompts to Pro. AI Plus manual selection changes depend on your account email. There is no verified universal subscription price increase.

Price Summary

Tool	What Changed	Was	Now	Impact	Who It Affects
Claude Sonnet 5.5	Direct API cache-read cut, October 7	$0.20/M cache-read tokens	$0.10/M cache-read tokens	Cache line -50%	Direct API workloads with cache hits
Claude Max / Team	Claimable API allowance announced October 7	Separately purchased usage in example budget	Max $100/$200; Team $20/$100 per seat, $500 pool cap	Conditional credit	Eligible existing subscribers using direct API
Gemini	Free rollout from October 9; Plus date is account-specific	Historical manual-selection matrix unverified	Free Auto; Plus manual Pro depends on notice	Access policy	Personal accounts, not Workspace

API amounts and credits are USD. One cache-read price cut, one included allowance and one access policy: these are different wallet effects. Current October 13 price captures do not establish October 5-11 price changes.

Past briefings: Week 40 · Briefing archive

Sonnet 5.5: a smaller cache-read line

On October 7, the direct Claude Platform cut Claude Sonnet 5.5 cache reads from $0.20 to $0.10 per million tokens. Input remains $2, output $10, five-minute cache writes $2.50 and one-hour cache writes $4 per million tokens. This is a cut to an existing model's cache-read rate.

For the same 100M cache-read tokens across requests, $20 becomes $10: a $10 saving on that component. An illustrative workload with 1M uncached input, 1M output and 100M cache reads costs $32 before and $22 now, excluding writes, tools and taxes. The whole invoice does not automatically halve.

Perplexity Agent still lists $0.20 for its Sonnet cache reads in the retained evidence. The direct-provider cut cannot be assumed on another platform. Haiku 5.5 is a separate new rate card, not a retroactive Haiku 4.5 cut.

Who This Affects

Developers already using direct Sonnet 5.5 with repeated cached context. Check cache-hit volume and the actual provider before changing models. See Claude API rates.

Official October 7 release notes and Claude Platform pricing. Sources retained October 13, 2026.

Max and Team: claim credits you can actually use

Anthropic announced included monthly direct API credits on October 7: Max 5x $100, Max 20x $200; Team Standard $20 per seat and Team Premium $100 per seat, pooled up to $500. Credits roll out over a few days and must be claimed. Subscription prices and usage limits are unchanged.

Your eligible plan must be active and in good standing for seven days. Link one Console organization. Max subscribers or Team Primary Owners/Owners need Console Owner, Admin or Billing permissions. Credits reset each billing cycle, monthly for annual plans, with no rollover. They are spent before purchased balance; without another balance, requests stop when the allowance is exhausted.

Free, Pro and Enterprise are excluded. Interactive Claude Code and extra Claude/Cowork usage are excluded, as are Amazon Bedrock, Google Vertex AI and Microsoft Foundry. An app or Agent SDK must bill eligible requests through the linked direct API key; subscription authentication alone does not use this allowance.

If you already pay $100/month for Max 5x and would separately buy exactly $100 of eligible API usage, claiming and fully using the $100 credit changes that illustrative budget from $200 to $100, excluding taxes and other discounts. This is a conditional saving, not a reconstructed historical invoice. Three Standard plus two Premium seats produce min($500, 3 × $20 + 2 × $100) = $260 of pooled credit.

Who This Affects

Existing eligible Max and Team subscribers building apps on the direct Claude API. Claim useful credits before buying more balance; do not upgrade solely for unused credits. See plans and separate API billing.

Official October 7 release notes, eligibility and claiming rules and plan prices. Sources retained October 13, 2026.

Gemini: check the account notice, then the model selector

Google's personal-account notice starts the Free model-selection rollout on October 9. Free Auto sends most prompts to Flash-Lite and can route deeper prompts to Flash or Pro. Turning smart selection off uses Flash-Lite. Free users have not lost every possible Pro answer.

The current help matrix lists Flash-Lite and Flash for AI Plus, with a cross for manually selected Pro. For Plus, the effective date is the date in your account email, not universally October 9. The announcement day and a complete historical before matrix are unverified.

Marketing cards and the help policy disagree on Plus access. Treat that contradiction as unresolved and check your own account. The current $4.99/month AI Plus and $19.99/month AI Pro comparison is $15 more per month only if you choose Pro for manual selection after your access changes. It is not a mandatory charge or a proved in-window list-price increase.

Who This Affects

Personal Gemini accounts, especially Plus subscribers who need manual Pro selection. Workspace is outside this notice. Try Free Auto and check the effective date before paying. See Gemini plan qualifications.

Official model-selection policy and Google AI marketing plans. Sources retained October 13, 2026.

Switch Watch

Three actions tied to actual billing and eligibility.

Existing API spend

Claim before topping up

If already eligible for Max or Team, link the right Console organization and claim credits you can use before their reset. Keep separate subscription and API budgets.

Cache usage

Check your provider and cache hits

Use the $0.10 direct Sonnet cache-read rate only for matching usage. Compare complete token costs and task results before moving workloads.

Account access

Check before upgrading Gemini

Read your Plus notice and try Free Auto. Pay the optional $15 monthly difference only if selectable Pro is needed and the current plan terms fit your account.

What is outside this week's price news

ChatGPT's October 7-8 Chat model rollout has no announced subscription price move and is not a price story. Cursor and Grok's October 12 signup bundle and the October 13 Gemini Argon capture are outside October 5-11. Current pricing corrections remain separate from dated weekly changes.

Reporting window: October 5-11, 2026 UTC. Evidence retained October 13. Unknown annual invoices, currency states and account eligibility are not inferred from other products or current marketing cards.