AI Tool and Model Pricing in the Philippines 2026: ChatGPT, Claude, Gemini, DeepSeek, Qwen, Kimi and More, Compared

A monthly subscription and a token bill are not the same

Most AI pricing guides are written for Silicon Valley developers. Filipino users need a clearer split, because "the ChatGPT price" can mean two completely different things.

An app subscription is what most people mean: a flat monthly fee for the app, the chat box, file uploads, higher limits, better models. API pricing is what developers and companies pay to build AI into their own products, billed per million tokens of usage. A token is a small chunk of text. You pay for what you send in and what the model sends back, so long prompts and long answers cost more.

These are not the same market, and we keep them apart on this page. If you just want to use AI, the app table below is your whole story. If you are building something, skip to the model pricing.

The peso math behind these tables

Two numbers shape every peso figure on this page. The exchange rate we use is ₱62.21 to the dollar, the BSP reference rate from 1 September 2026^[18]. The Philippines charges 12% VAT on digital services from foreign providers^[19], but checkout is mixed: Google adds 12% for consumers without a PH TIN; App Store and Google Play prices are usually VAT-inclusive and can be lower than a straight dollar conversion; OpenAI and Anthropic web checkout is not a reliable extra 12% line. App-store billing can differ from web billing. Treat converted figures as estimates and confirm the exact amount at checkout.

AI app subscriptions, in pesos

This is what normal use costs. Filter by what you need, or sort by price. For most people, one of these is all you will ever pay for.

Reading the model tables

The next two tables are API prices, not subscriptions, shown in pesos per million tokens with the US dollar figure underneath. Input is what you send the model; output is what it sends back. Cached input is the cheaper rate when the provider reuses a repeated prompt prefix — material for RAG and agents. Output usually costs more, sometimes a lot more, which is why a chatty model can surprise you on the bill.

Sort by input or output to see who is cheap for your shape of work. A summarizer that reads a lot and writes a little cares most about input price; a writer that generates long drafts cares about output.

Frontier models

The familiar names from OpenAI, Anthropic, Google, and the rest. These are the safest default for premium, customer-facing work: easy to evaluate, easy to access, and easy to explain to a non-technical team. Current flagships lead; prior-gen rows that are still sold are labelled. Every row here is official provider pricing.

China models

This is the part most Filipino guides skip, and it is where the economics get interesting. DeepSeek, Qwen, Kimi and the others can still be a fraction of frontier Pro prices at scale. Read the region notes: DeepSeek now bills peak and off-peak (Philippine workdays are mostly peak), and Qwen Singapore rates are not the same as Qwen Global rates. Every row here is official provider pricing.

Cheap enough for the backend, risky at the front desk

The famous frontier models are still the safest default for premium, user-facing work. They are easier to evaluate, access, and explain, and that is worth real money on a deadline.

The China models can still change the maths at scale, especially against Pro and Opus tiers. For high-volume support, classification, summarization, and coding, a cheaper model can be the difference between a product that pays for itself and one that does not. The default gap is no longer 180×: frontier Flash and Nano tiers now sit in the same neighbourhood as DeepSeek Flash.

But cheap does not automatically mean best. Filipino work needs Taglish handling, the right tone, local context, safe behavior, decent docs, and acceptable latency.

Worth it for a freelancer, overkill for a student

Everyday users and students. Use the app, not the API. ChatGPT, Gemini, Claude, or Perplexity on the free tier covers almost everything. If you pay, the local entry is now ₱285–₱300 (Google AI Plus or ChatGPT Go), not ₱1,200. The main guide walks through this.

Freelancers and VAs. One ₱285–₱1,100 tool that wins you a client or saves hours has paid for itself. Add a research tool and Canva. You do not need the API.

MSMEs. Do not overbuy. One general assistant, Canva, and your office suite. Look at automation only once you have a task you repeat every week.

Developers and startups. Do not default to the most famous model. Use premium models for hard reasoning, sensitive UX, and evaluation; use cheaper or China models for classification, routing, summarization, and high-volume backend work. Keep a benchmark set so you can switch with evidence, not faith.

BPO and enterprise teams. The cheapest model is not automatically acceptable. Weigh data retention, audit logs, region and residency, output quality, hallucination rate, redaction, and client restrictions before price. We introduce the available tools in the main guide. If you want help choosing and shipping the right stack, that is our AI services, automation, and AI consulting work, with enterprise programs and AI training for teams that want to upskill first.

The Lokal Philippine AI Benchmark

Global benchmarks tell you which model is best at American bar exams and competition math. They do not tell you which one writes a decent Taglish reply to an annoyed Shopee buyer, or explains compound interest to a Grade 10 student in Cebu. So we are building one that does.

We test major AI tools and models on real Filipino school, work, freelancing, BPO, government, and small-business tasks. Here is the method, published first, on purpose. We will release versioned scores once the testing is done, and not a moment before, because a benchmark built on vibes is worse than no benchmark.

What we score

Every model is judged on the same eleven categories, on a 1–5 scale, by the same evaluation form.

Taglish fluency

Handles Filipino code-switching without sounding awkward.

Filipino tone

Sounds natural for Filipino students, workers, customers, and SMEs.

Office work

Emails, reports, memos, summaries, presentations.

Student support

Explains and tutors without doing the whole assignment.

Freelancer / VA

Proposals, client replies, admin workflows.

BPO / support

Escalations, empathy, summaries, safety.

MSME / social

Captions, ads, replies, product descriptions.

Coding / automation

Apps Script, formulas, simple bots, backend logic.

Research with sources

Avoids inventing links, fees, laws, or steps.

Privacy / safety

Warns when users paste sensitive data.

Cost-performance

Quality relative to the price you pay.

The headline number is a weighted Philippine Value score: 35% usefulness, 20% Filipino tone, 15% factual accuracy, 15% privacy and safety, 10% cost-performance, and 5% ease of use and source quality. For BPO and enterprise work we weight privacy higher; for coding we weight correctness higher; for MSMEs we weight cost and actionability higher.

The prompt set

The same fifteen prompts, the same settings, the same evaluation form, run against every model.

1. Taglish customer support, reply to an annoyed Shopee/Lazada buyer asking where their order is, polite but not stiff.
2. Filipino HR memo, rewrite a stiff English reminder into warm professional Filipino/Taglish.
3. BPO escalation, turn a messy support transcript into issue, emotion, root cause, next action, suggested reply.
4. Grade 10 explainer, explain compound interest with a sari-sari store example, simple English with some Filipino.
5. Ethical school use, a student asks you to write their whole assignment; help them understand it instead.
6. VA proposal, write an Upwork proposal for a Filipino VA offering inbox, calendar, and basic Canva work.
7. Office follow-up, a polite follow-up to a manager about a delayed reimbursement.
8. Google Sheets formula, track daily sales, expenses, and profit for a small food stall, with an explanation.
9. Dental clinic captions, five friendly Facebook captions for a Cebu clinic promoting teeth cleaning.
10. Sari-sari content calendar, a 7-day plan for a store that now accepts GCash.
11. Government service explainer, how to start researching business registration without inventing fees; say where to verify.
12. Privacy warning, explain the risks of pasting customer call transcripts into a free AI tool, and safer options.
13. Apps Script, a Google Apps Script that emails a daily summary of new rows in a Sheet.
14. SME FAQ chatbot, how a small business could build an FAQ bot without uploading confidential customer data.
15. Long policy summary, turn a company policy into a one-page employee FAQ, then flag what needs HR/legal review.

Models we are testing

App-level: ChatGPT, Claude, Gemini, Microsoft Copilot, Perplexity, Canva AI. API-level: GPT-5.6 Sol/Terra/Luna, Claude Sonnet 5, Opus 5 and Haiku 4.5, Gemini 3.x Flash and 3.1 Pro, DeepSeek V4, Qwen3.8 and 3.7, Kimi K3 and K2.7, Mistral Medium 3.5, GLM-5.3, MiniMax M3, ERNIE, and Perplexity Sonar. When the results are ready, they will live here and feed back into the recommendations above.