DEV Community: MUHAMMAD ZAIN UL ABIDIN

Freelancing from Pakistan as a Web Developer in 2026: The No-Fluff Playbook

MUHAMMAD ZAIN UL ABIDIN — Wed, 14 Oct 2026 05:20:52 +0000

Pakistani freelancers pulled in a record $1.76 billion in foreign exchange
during FY26, up 78% from the year before, with IT freelancers contributing
$1.16 billion of it. Nearly three million Pakistanis now freelance, making
the country the world's fourth-largest freelance market. This post is the
no-fluff version of how a web developer actually gets into that number:
which platform to pick, what the fees really cost, how the 0.25% tax works,
how to get paid, and what to do in your first 90 days.

The short answer

Start freelancing as a web developer from Pakistan in 2026 by picking one
niche (Next.js builds, e-commerce, or AI integrations), building two or three
portfolio projects that look like paid work, opening on Fiverr (free to start,
20% commission) or Upwork (variable 0-15% fee, ~10% typical, plus $0.15
Connects per bid), registering with PSEB to qualify for the 0.25% final tax
rate on IT export income, and receiving payments through Payoneer or Wise.
Expect your first client in 30-90 days of daily effort.

Why 2026 is genuinely a good year to start

Most "start freelancing" guides lean on vibes. The State Bank of Pakistan
numbers do not need any:

$1.76B freelance earnings in FY26 (State Bank of Pakistan data, reported by Dawn, The News, TechJuice, July 2026)
78% year-on-year growth (Up from $984 million in FY25 - SBP figures)
~3M freelancers in Pakistan (Industry estimates (PAFLA) - fourth-largest freelance workforce in the world)

The breakdown matters more than the headline. IT freelancers generated $1.16
billion (up 49%), while non-IT freelance earnings tripled to $592 million -
writing, design, marketing, and virtual assistance. Two readings of that: the
demand for technical work keeps compounding, and clients are buying wider
skill bundles than before. A web developer who can also ship the landing
page, wire up the analytics, and write the launch copy sells a bigger ticket
than one who ships code alone.

Where these numbers come from
Freelance remittance figures are from State Bank of Pakistan data released
in July 2026 and reported by Dawn, The News, Pakistan Today, TechJuice,
and ProPakistani. The 0.25% tax extension comes from coverage of the
Finance Act 2026 by Business Recorder, TechX, and PhoneWorld. Platform fee
figures are 2026 published rates; confirm the current number on each
platform's official fee page before you quote it to yourself, because these
change without notice.

Pick your platform: the 2026 fee reality

Your platform choice is a tax on your income, so do the math before you fall
in love with anyone's marketing. Here is what each major platform actually
costs a Pakistani web developer in 2026:

Platform	Freelancer fee	Bidding cost	Best for
Fiverr	Flat 20% on every order	None (gig-based)	Zero-capital starts, packaged services
Upwork	Variable 0-15% per contract (most report ~10%)	$0.15 per Connect, ~2-16 Connects per proposal	Custom projects, long-term clients
Freelancer.com	10% (or $5 minimum)	Free bids on plan tiers	Contest-style and budget work
Toptal	0% to the freelancer	Screening, not bidding	Senior devs who can pass the vetting

Two practical conclusions. First, Fiverr's 20% is the highest headline rate
in the table, but it costs you nothing before you earn, which makes it the
only rational first platform when you are bootstrapping from zero. A $100
order nets you $80 from day one with no ad spend. Second, Upwork's effective
rate is lower than Fiverr's for most freelancers, but the Connects system
means you spend $3-5 per proposal with no guarantee of a reply. If you burn
through 50 Connects a week at $0.15 each with weak proposals, Upwork is a
hole you pour money into.

The right sequence for most Pakistani beginners: Fiverr first for the first
client and first reviews, Upwork second once you can afford to bid on ten
jobs a week, and direct clients third once you have a portfolio worth
showing off-platform. Never build your entire income on one platform - an
account flag, a fee change, or an algorithm shift can take 100% of your
revenue overnight.

The six things you need before your first client

Nothing on this list costs more than a few evenings. Every item you skip
becomes a delay later, so treat it as a checklist, not advice.

One niche, not five

"Web developer" is not a niche, it is a commodity listing. Pick one: Next.js
marketing sites, Shopify/WooCommerce builds, SaaS MVPs, or AI integrations
(chatbots, RAG, MCP tools). Your niche decides your gig titles, your
portfolio, and which jobs you bid on. Everything below compounds from this
choice.

Three portfolio projects that look like paid work

Build three projects in your niche as if a client paid for each one: real
brief, real deadlines, live deployment link, GitHub repo, and a short
case-study writeup explaining the problem and your decisions. Clients do not
read code on a first pass, they look at the live thing and the writeup.
Three strong projects beat ten tutorial clones every time.

One platform profile done properly

Profile photo (real face, decent light), headline with your niche and one
outcome ("Next.js sites that load in under 2s"), a description written for
buyers not recruiters, and your portfolio projects linked. On Fiverr, build
2-3 gigs with tiered packages (Basic/Standard/Premium). On Upwork, pick 10-15
skills that match your niche exactly. Generic profiles get generic results.

One payment rail, verified with a small withdrawal

Set up Payoneer or Wise (Elevate Pay is worth comparing as a
Pakistani-built option) before you need it, and run one small test
withdrawal to your local bank. Discovering your payout path is broken the
day a $500 client payment lands is a special kind of panic you only need
once. PayPal is still not functional in Pakistan, so do not plan around it.

PSEB registration on your 90-day list

Registration with the Pakistan Software Export Board is what qualifies your
IT and IT-enabled service export income for the 0.25% final tax rate under
Section 154A, instead of normal income-tax treatment. It is a government
portal process, not a lawyer's job. Do it once your income is real and
before filing season, not after.

A written proposal template, not a begging letter

Your template: one line naming the client's actual problem, two lines on
how you would solve it, one line of proof (a link), and the price. Custom
beats clever every time. Send 3-5 tailored proposals a day on Upwork or
answer every Fiverr inquiry within the hour. Volume of quality attempts is
the whole game at the start.

What web developers actually charge

Rate anxiety kills more freelancing careers than competition does. The
numbers below are indicative ranges seen across platforms for Pakistani
developers in 2026 - not guarantees, and your niche, reviews, and how well
you sell all move you up or down the band:

Service	Starter range	With reviews + niche
Landing page / marketing site	$100-300	$500-1,500
Full website (Next.js / WordPress)	$300-800	$1,000-3,000
E-commerce build	$500-1,200	$1,500-5,000
SaaS MVP	$1,000-2,500	$3,000-10,000+
Hourly dev work	$15-30/hr	$40-80/hr
AI integration / chatbot build	$400-1,000	$1,500-4,000

Price for take-home, not for the platform
These are gross figures. Run every quote through your platform's cut before
you accept it: a $1,000 Fiverr order is $800 in your account, and a $1,000
Upwork contract at a 10% fee is $900 minus whatever Connects you spent
winning it. The section below works the math on a real example so you stop
doing it in your head mid-proposal.

The money math: what a $1,000 project leaves you

Concrete example, because fee percentages stay abstract until you lose real
money to them. You win a $1,000 fixed-price website build:

On Fiverr: 20% commission = $200 to the platform, $800 to you. No bidding cost, 14-day clearance hold before withdrawal.
On Upwork: ~10% service fee = ~$100 to the platform, ~$900 to you, minus Connects. Say you spent 12 Connects ($1.80) across two rounds of bidding to win it: you keep ~$898, minus withdrawal fees depending on your method.

Then the tax layer: if you are PSEB-registered and the income qualifies as
IT export proceeds under Section 154A, the 0.25% final tax on $900 is $2.25,
and that is the whole tax bill on that income. Without PSEB registration,
the same income can fall under normal tax treatment - a difference of
hundreds of dollars on a $1,000 project. This is why the paperwork section is
not bureaucratic decoration.

The lesson is not "Upwork always wins." It is that Fiverr's simplicity costs
about $100 per $1,000 project versus Upwork's cheaper-but-pay-to-bid model,
and that the tax setup dwarfs both. A freelancer who fixes their tax
registration earns more from a Fiverr order than one who ignores it earns on
Upwork.

The tax setup nobody explains properly

Pakistani freelancers get two pieces of tax news in 2026, one good and one
demanding attention. The good: the Finance Act 2026 extended the 0.25% final
tax rate on IT and IT-enabled service export income for PSEB-registered
exporters through June 30, 2029. It is a final tax - once it is deducted
from your export proceeds, you owe nothing more on that income.

The part that needs attention is qualification. The 0.25% rate applies to
export proceeds from computer software, IT, or IT-enabled services where the
exporter is registered with and certified by PSEB, received through proper
banking channels. Receiving money from a foreign client does not
automatically make every service PSEB-qualified, and non-filers face higher
withholding on top. Keep client invoices, bank records, and your PSEB
registration in one folder. Filing season runs on its own calendar - the
September 30 deadline catches first-timers every year.

Register with PSEB, get paid through banking channels, file your return.
Everything else about freelancer tax in Pakistan is detail around that sentence.

-- The one-sentence version

Getting paid: pick one rail and verify it

Three rails matter for Pakistan in 2026. Payoneer is the default most
platforms integrate with natively. Wise usually wins on exchange rates for
direct client transfers. Elevate Pay is the Pakistani-built option designed
for cheaper direct transfers and keeps your dollars in a foreign-currency
account with local withdrawals. Each has different fee structures and payout
speeds, and the cheapest rail for a $50 gig is not always the cheapest for a
$5,000 contract - that comparison deserves its own post, and it is coming.

For now, the rules: one primary rail plus one backup, a test withdrawal
before you need it, and never your entire balance sitting on a platform.
Withdraw on a schedule, keep payment screenshots with your invoices, and
remember every withdrawal is also your tax documentation.

Your first 90 days, week by week

Days 1-14: build the machine

Pick your niche, finish one portfolio project end-to-end with a live URL and
a writeup, set up your payment rail, and complete your platform profile.
No proposals yet. A half-built profile sending proposals is how beginners
burn their first 100 Connects for nothing.

Days 15-45: ship and propose daily

Finish portfolio projects two and three. Start sending 3-5 tailored
proposals a day on Upwork, or launch your Fiverr gigs and share them where
your buyers are (LinkedIn, relevant communities - not spam, answers). Target
small jobs first: clients hire beginners for $100-300 test projects, not
$5,000 builds.

Days 46-90: convert, deliver, repeat

Your goal is 3-5 completed jobs with 5-star reviews, not one big win. Over-
deliver on the first ones - fast communication, clean handoff, a short Loom
walkthrough. Each review multiplies your next proposal's reply rate. Start
your PSEB registration the week your second client pays.

The fluff version:

"Just create an Upwork profile and wait for clients!"
"Charge $5/hour to get started!"
"Taxes? Worry about that when you're rich."
One generic profile, copied proposals, rage-quitting at week six.

The version that works:

Niche down, build three projects that look like paid work.
Price at the starter band of your niche, raise every 5 reviews.
PSEB registration + banking channels + invoices from the second client.
Daily proposals, weekly shipping, 90 days of patience.

FAQ

How much can a Pakistani web developer earn from freelancing?
Ranges vary by niche, but Pakistani web developers on international platforms
commonly report $15-40/hour starting out and $50-150/hour with specialization
and reviews. In FY26, Pakistani freelancers earned a record $1.76 billion,
with IT freelancers contributing $1.16 billion, so the demand side is real -
your earnings depend on niche, positioning, and how consistently you ship.

Is Upwork or Fiverr better for beginners in Pakistan in 2026?
Fiverr costs nothing to start: no bidding fees, you build gig listings and
clients come to you, but you pay a flat 20% commission on every order. Upwork
charges a variable 0-15% fee per contract (most report around 10%) plus $0.15
Connects to bid, so it costs money before you earn. Start with Fiverr if you
have zero capital, add Upwork once you can afford to bid strategically.

Do I need PSEB registration to freelance legally from Pakistan?
You can start freelancing without it, but registering with the Pakistan
Software Export Board is what unlocks the 0.25% final tax rate on IT and
IT-enabled service export income under Section 154A. Without registration,
the reduced rate does not apply to you, so the tax difference is large enough
that registration should be on your 90-day list.

What is the freelancer tax rate in Pakistan in 2026?
For PSEB-registered IT and IT-enabled service exporters, the rate is 0.25%
final tax on eligible export proceeds received through banking channels,
extended by the Finance Act 2026 through June 30, 2029. Other export income
under Section 154A is taxed at 1%. The 0.25% rate is final - you owe nothing
more on that income.

How do Pakistani freelancers receive money from foreign clients?
The common rails are Payoneer, Wise, and Elevate Pay, with bank withdrawals
into local accounts. PayPal still does not work properly in Pakistan, so do
not plan around it. Pick one rail, verify it with a small withdrawal before a
big one, and keep client payments and personal spending separated for clean
tax records.

How long does it take to get your first freelancing client?
With daily effort - 3-5 quality proposals a day on Upwork or a few well-built
Fiverr gigs - most beginners land a first small project in 30-90 days. The
first client is the slowest because you have no reviews; after five completed
jobs with ratings, reply rates climb sharply. There is no shortcut here that
does not involve sending proposals and shipping work.

The tools worth bookmarking

Tool	What it does	Link
Fiverr	Gig-based marketplace, zero cost to start	https://www.fiverr.com
Upwork	Proposal-based marketplace for custom projects	https://www.upwork.com
Payoneer	Receive international payments, local bank withdrawal	https://www.payoneer.com
Wise	Multi-currency transfers, usually best exchange rates	https://www.wise.com

Want the freelancer playbook applied to your project? If you are a
founder who needs a Next.js site, a store, or an AI integration shipped by
someone who understands both the code and the business math, reach out
through my about page. I take on a small number of projects at a
time, and I answer every serious inquiry personally.

Claude Haiku 5.5 in Claude Code: Cut AI Coding Costs 20x

MUHAMMAD ZAIN UL ABIDIN — Tue, 13 Oct 2026 05:26:35 +0000

Anthropic released Claude Haiku 5.5 on October 7, 2026, and the headline
number is hard to ignore: $0.10 per million input tokens, a tenth of Sonnet
5.5 and a fortieth of Opus 5.5. The real question for anyone running Claude
Code daily is not the price, though. It is which jobs a $0.10 model can take
away from a $2 model without your pull requests noticing. I dug through the
launch benchmarks, the independent coding tests, and the pricing fine print
so you can route your work before your next invoice surprises you.

The short answer

Claude Haiku 5.5 is Anthropic's cheapest, fastest model, and the first Haiku
that can survive an agent loop: 1M-token context, adaptive thinking with
adjustable effort, and agent-benchmark scores several times higher than Haiku
4.5. Route narrow, verifiable work (subagent lookups, extraction, summaries,
one-shot fixes with tests) to Haiku at $0.10/$0.50 per million tokens, keep
Sonnet 5.5 as your default coder, and watch the 100K-token price cliff that
re-prices the whole request 5x when you cross it.

The price, verified

Every number below is per million tokens, in USD, from Anthropic's own
pricing for the Claude 5.5 family. I am leading with the vendor's table
because every independent benchmark prices against it:

Model	Input / output (per 1M)	Cache read	Context
Haiku 4.5	$1 / $5	$0.10	200K
Haiku 5.5 (up to 100K prompt tokens)	$0.10 / $0.50	$0.01	1M
Haiku 5.5 (over 100K prompt tokens)	$0.50 / $2.50	$0.05	1M
Sonnet 5.5	$2 / $10	$0.10	1M
Opus 5.5	$4 / $20	$0.20	1M

Two things jump out. First, the gaps are clean multiples: Haiku is 20x
under Sonnet and 40x under Opus on every token class. Second, Haiku 5.5 is
the only model in the lineup with two price tiers, and the tier boundary is
a cliff, not a slope.

Benchmark figures in this post come from Anthropic's launch materials
(vendor-reported, not independently reproduced) and from Bito's October
2026 coding benchmark, which ran fresh sessions on its own task battery.
Pricing was cross-checked against Anthropic's pricing page as reported on
October 11, 2026. Treat vendor benchmarks as directional, not gospel.

Anthropic itself puts the average saving at about 75%, not the 90% the
headline implies - the gap closes once long prompts, the new tokenizer, and
output tokens enter the math. That is still a big number. It is just not the
number on the poster.

What it can actually do now

The reason this Haiku matters is the capability jump, not the price cut.
On Anthropic's own table, Haiku 5.5 beats Haiku 4.5 on every row where both
have a score, and the deltas on agentic benchmarks are the story:

Benchmark	Haiku 4.5	Haiku 5.5
OSWorld 2.1 (computer use)	15.7%	72.4%
Terminal-Bench 4.0 (shell tasks)	0.0%	39.2%
GDPval-AA v2.1 (professional tasks, Elo)	735	1620
Humanity's Last Exam, no tools	10.2%	45.9%

The honest reading: this is a genuinely capable small model for
structured agent work, and it is still not a frontier coder. The
hardest line in the whole table is Terminal-Bench: 39.2% against Sonnet's
70.6%. Multi-step shell work, large refactors, debugging across a codebase -
Anthropic itself points those at Sonnet 5.5 and Opus 5.5. Haiku 5.5 takes
narrow, well-scoped edits. It does not take open-ended engineering.

The more practical evidence comes from Bito's benchmark, which ran real
coding sessions the week of launch. On one-shot asks, Haiku 5.5 at high
effort scored 87% at $0.014 per session against Opus 5.5 at 90% for $0.32 -
three points of quality for roughly 1/23rd of the cost. On long build
sessions the gap widens: Haiku at high effort hit 82% for $1.35 while Opus
took 85% for $5.42. That is the shape of this model. The simpler and more
checkable the job, the more of the price gap you keep.

The 100K-token cliff nobody reads

Here is the pricing detail that will actually change your bill. Haiku 5.5's
$0.10/$0.50 rate applies to prompts of 100,000 tokens or fewer. Cross that
line and the entire request reprices at $0.50/$2.50 - not just the tokens
above the line, the whole thing, cache reads and cache writes included.
Each request is priced on its own.

Work the example: an agent session with 120,000 input tokens does not cost
the cheap rate for the first 100K and the expensive rate for the last 20K.
It costs the expensive rate for all 120K. At $0.50 input per million, that
is $0.06 instead of $0.012 - a 5x jump from a 20% overage. Agent loops that
keep stuffing conversation history back into the prompt cross this line
quietly, session after session.

Two defenses. First, context discipline: compact or summarize history
before it approaches 100K, and keep the system prompt lean. Second, the
new tokenizer counts roughly 30% more tokens for the same text than Haiku
4.5's did (Simon Willison measured about 1.25x on a long test prompt with
his token counter), so a prompt that fit comfortably under the old model
may not fit under the new one. Run your own prompts through a counter
before planning a migration.

The Checkability Rule: a routing framework

Forget "big model good, small model bad." The routing question has a
precise form, and I am going to name it because it is the decision rule I
actually use:

The Checkability Rule: route a task to Haiku 5.5 when you can verify
its output for less than the price gap between the models. Route to
Sonnet or Opus when you cannot.

Verification is the whole game. A test suite, a type checker, a linter, a
schema validator, a human glance at a diff - these are all cheap oracles.
When the oracle is cheaper than the price gap, the small model's mistakes
are free: you catch them and retry, and you are still ahead. When there is
no oracle - ambiguous product decisions, novel architecture, judgment
calls - the price gap buys you nothing and the failure costs you hours.

In practice this produces a stable split:

- Subagent lookups: search, read, summarize for an orchestrator - Classification, routing, extraction, tagging - Summaries, changelogs, context compaction - One-shot fixes with an existing test suite - Boilerplate generation you will diff anyway - Anything with a cheap oracle: tests, types, schemas - Open-ended feature builds and architecture calls - Multi-step shell and refactor work (the 39.2 vs 70.6 gap) - Debugging across an unfamiliar codebase - Anything where a wrong answer is expensive - Long-horizon autonomous agent runs - Tasks with no acceptance test you trust

I write my acceptance checks as typed contracts - Zod schemas as the
contract between the agent and me - because a
check you can run in CI is a check you can route against. The sharper your
oracles, the more work clears the bar for Haiku.

One more routing lever: effort. Haiku 5.5 ships with adaptive thinking and
five effort levels, defaulting to medium. Use low for routing and simple
labels, medium for most extraction, and high for multi-step checks
where a wrong call is expensive. Bito's numbers show effort is real money:
Haiku at high effort beat Sonnet's default on every task set they measured,
at roughly half the session cost.

Setup in Claude Code

Haiku 5.5 needs Claude Code v2.1.293 or later for full support. Run
claude --version and update if you are behind. Older versions may not
list the model or may mishandle its thinking configuration.

Inside any session, run /model claude-haiku-5-5 to switch the current
session to Haiku. The model ID is claude-haiku-5-5 everywhere - the
API, Amazon Bedrock (anthropic.claude-haiku-5-5), Google Cloud, and
Microsoft Foundry all use the same string.

For sessions you know are narrow - triage, exploration, bulk labeling -
launch directly with claude --model claude-haiku-5-5. Starting on
Haiku beats switching later, because the model never pays the
orchestrator's context prices for the opening turns.

Keep your main session on Sonnet (or Opus for hard jobs) and let it
spawn Haiku subagents for the lookup and verification work. Measured
spawn costs make the case: a fresh subagent prefix costs roughly
$0.005-0.011 on Haiku 5.5 against about $0.17 on Sonnet 5.5. That is the
cheapest architectural change in this whole post.

Start at the medium default, drop to low for routing and labeling
where speed matters most, and raise to high for the multi-step checks
where a wrong call is expensive. Effort is the second pricing lever
after model choice, and it costs nothing to experiment with.

Where Haiku still loses: the honest list

A cheap model you trust blindly is worse than an expensive one, so here is
everything that tempers the headline:

The coding gap is real. 39.2% vs 70.6% on Terminal-Bench is not a matter of degree. Open-ended engineering belongs to Sonnet and Opus, and Anthropic says so on its own launch page.
The 100K cliff. Covered above, but it bears repeating: one long context window wipes out the per-token savings for the whole request.
The tokenizer tax. About 30% more tokens for the same text than Haiku 4.5, measured independently. Your old cost math needs re-running.
Not a drop-in replacement. Manual budget_tokens thinking, non-default sampling parameters, and assistant prefill now return errors. Computer use needs a new toolset version. Budget migration time for anything built on Haiku 4.5's quirks.
Effort at max writes more. Independent analysis notes Haiku 5.5 at max effort emits roughly 3x the output tokens per task as its closest competitor, which eats into the per-token price advantage on output-heavy work.
Vendor benchmarks are marketing. Anthropic's table compares against competitors' reported figures; the methodology details live in a system card most people will not read. Independent coverage is still thin - this model is six days old. Expect the numbers to move.

What this does to your bill

Take Bito's one-shot numbers as a worked example, since they are the
closest thing to a real workload. Suppose your week holds 200 one-shot
coding asks:

Setup	Score	Cost per ask	Weekly cost
Opus 5.5 default	90%	$0.32	$64.00
Haiku 5.5 high effort	87%	$0.014	$2.80

Same week, same asks: $64 becomes $2.80 for three points of benchmark
score, provided every ask is checkable. That "provided" is doing all the
work, which is why the Checkability Rule is the post and the table is just
illustration. Anthropic's own guidance lands in the same place: they quote
roughly 75% average savings in practice, not 90%.

The compounding win is the orchestrator pattern. A Sonnet session that
spawns ten Haiku subagents a day for lookups, each costing about a cent
instead of $0.17, saves roughly $1.60 a day on spawns alone - before any of
the per-token math. Small numbers, every day, automatic.

Tools for the routing workflow

{/* MONETIZATION NOTE (not rendered): the links below are official product
pages. If Anthropic or a benchmark provider opens an affiliate program,
swap in affiliate links here and update the disclosure. Never claim
affiliate status before it exists. */}

Tool	What it does	Why it matters here
Claude Code	Anthropic's agentic coding CLI	Where you run `/model` and the orchestrator pattern
Anthropic API	Usage-based API access	Direct per-token billing when subscriptions don't fit
Artificial Analysis	Independent model benchmarks	Tracks whether Haiku's scores hold as tests mature

The links above are plain official links, not affiliate links. If that
changes I will say so on the page, same policy as every post here.

FAQ

What is Claude Haiku 5.5 and when was it released?

Claude Haiku 5.5 is Anthropic's small, fast, cheap model in the Claude 5.5
family, released October 7, 2026. It brings a 1M-token context window,
adaptive thinking with effort levels, and benchmark scores that beat Haiku
4.5 by wide margins on agentic tasks.

How much does Haiku 5.5 cost in Claude Code?

Through the API it costs $0.10 per million input tokens and $0.50 per
million output tokens for prompts up to 100,000 tokens, jumping to
$0.50/$2.50 above that line. In Claude Code subscription plans, model
choice does not change the subscription price, but routing work to Haiku
keeps you under usage caps and stretches rate limits further.

How do I use Haiku 5.5 in Claude Code?

Inside a session, run /model claude-haiku-5-5 to switch models. To start
a session on Haiku directly, launch with claude --model claude-haiku-5-5. You will need Claude Code v2.1.293 or later for full
support. The model ID is claude-haiku-5-5 everywhere: API, Bedrock, and
Vertex.

Is Haiku 5.5 good enough for coding agents?

For narrow, checkable coding jobs, yes - independent benchmarks put Haiku
5.5 at high effort within a few points of Opus 5.5 on one-shot asks at
about 1/23rd the cost. For open-ended multi-step engineering it still
lags: Anthropic's own figures show 39.2% vs 70.6% on Terminal-Bench 4.0,
which is why Sonnet stays the default for everyday coding.

What is the 100K-token price cliff?

Haiku 5.5's headline price only applies to prompts of 100,000 tokens or
fewer. Cross that line and the whole request - input, cache reads,
everything - reprices at $0.50/$2.50 per million tokens, five times higher.
Long agent histories and repo-wide context cross the line easily, so keep
prompts compact or pay the second tier.

Should I switch from Sonnet 5.5 to Haiku 5.5?

Not wholesale. Use Sonnet as your orchestrator and default coder, and
route narrow, verifiable work - classification, extraction, summaries,
subagent lookups - to Haiku. That hybrid setup is where the savings live.
Switch the default only for workloads where Haiku's scores already clear
your acceptance bar.

Cut your agent bill on purpose

I help teams and founders set up exactly this: agent workflows where the
model routing, the acceptance checks, and the cost monitoring are designed
together instead of discovered in the invoice. If your Claude Code or API
bill is growing faster than your output, that is a solvable problem, and
it is usually a week of work, not a rewrite. Details are on my
about page.

Route by checkability: if you can verify it cheaply, Haiku does it. If
you cannot, that is what Sonnet is for.