llm-pricing on S Anand

LLM Deprecations and Price Changes

Thu, 21 May 2026 14:02:10 +0800

A colleague told me a near-miss horror story.

As Google began deprecating Gemini 2.0, we moved to Gemini 2.5 Pro. But reasoning is enabled by default and cannot be turned off.

For our specific problem statement, reasoning was not required. Token costs increased 10x and speeds were 3-4x slower. We moved the client to Gemini 2.5 Flash Lite, which has reasoning turned off by default and offers much lower latency.

Because we track compute costs closely, we managed this without a major financial impact.

But model updates clearly require careful testing on the cost and latency front as well, not just output quality.

AI used to keep getting cheaper. But now it’s more a “convergence”.

Each line traces a model family. The X-axis is its intelligence (LMArena ELO score) and the Y-axis is the input cost ($ per million tokens, log scale). Time flows roughly left to right as models improve.

Three patterns emerge and each seems strategic.

Gemini Flash rockets upward (cheap -> expensive, moving right).
GPT-4 class collapses from $30 toward $1.25, then GPT-5.5 jumps back up to $5.
Claude Sonnet runs perfectly flat (cheap stays cheap as it gets smarter).

Google’s Loss Leaders

Gemini 1.5 Flash was $0.075/MTok in August 2024. I told everyone to use it - it’s a fantastic, affodable model.

Gemini 3.5 Flash, released this week at Google I/O, is $1.50. That’s a 20× in 21 months. Looks like “Flash” is more a brand position than a price position. It migrates up-market.

Same for Gemini 1.5 Flash 8b (3.75 cents) which migrated into Gemini 3.1 Flash Lite (25 cents - a 6.7× increase).

Gemini Pro went the opposite direction: down from $5.00 (1.5-pro in Oct 2024) to $1.25-$2.00 today. Pro seems to be a competitive weapon against Anthropic and OpenAI at the top, while monetizing the middle.

Of course, Gemini’s real lock-in is the Google Workspace bundling and Search AI Mode. Personally, I subscribed to Google Pro for the first time in 20 years just for these bundled capabilities.

By the time people notice the Flash price, it’s hard to leave the ecosystem.

OpenAI’s Relaunches

The price chart speaks for itself:

Model	Price
GPT-4	$30
GPT-4 Turbo	$10
GPT-4o	$2.50
GPT-4.1	$2.00
GPT-5	$1.25
GPT-5.2	$1.75
GPT-5.4	$2.50
GPT-5.5	$5.00

Maybe their strategy seems is: scale current technology to commodity (the O-series showed the same pattern with $15 for O1 falling to $1.10 for O4-mini in 7 months), THEN launch a new frontier above it, and repeat?

Anthropic’s Revisions

Claude Sonnet has held at exactly $3.00/MTok input for over two years, across four model generations and an ELO gain of nearly 200 points. Pretty unusual.

Opus came down from $15 to $5 in November 2025 - likely a deliberate move to make it production-viable. Haiku crept from $0.25 to $0.80. The tiers are converging.

There’s a twist, though. Anthropic restructured enterprise contracts in late 2025. Seat prices dropped to $20/seat. Token discounts (previously 10-15% off API rates) were removed. For a 100-seat team, that adds ~$15K-$40K to annual TCO. So, prices went down, but the actual bill went up.

Also, Opus 4.7 uses a new tokenizer that may consume ~35% more tokens for the same text. It’s worth re-benchmarking prompts before assuming $5 is 1/3rd of $15.

What Do We Do?

Model family prices change rapidy. Old models get deprecated. Best to be prepared.

Add multi-tier routing to your architecture:
- Lite / Nano / Haiku for extraction, classification, … – tasks with clear right answers
- Sonnet / GPT-5.4 / Gemini Flash for most production reasoning
- Opus / GPT-5.5 for escalation or expert advice: complex planning, hard edge cases, high-value decisions
Compare completed-task costs, not token price. A 2* more expensive model can halve the retry rate, making it cheaper per successful output.
Migrate by model capability, not model family. Switch to models with similar latency, context window, output format compliance and reasoning depth.
Evaluate open-source models. DeepSeek models at self-hosted inference costs can be 90% cheaper for commodity (not frontier) tasks.

For enterprise procurement: keep a close eye on pricing changes, API token discounts, and what was quietly removed during renewal. (AI helps with this!)

The first 18 months of most AI model families are discounted customer acquisition. Then value extraction follows. Google started it with Flash. OpenAI is doing it with GPT-5.5. Anthropic is doing it with enterprise billing restructuring.

Fair enough. Providers need to recover infra investments.

And build good routing strategies helps enterprises get the most out of this.

Just keep asking yourself: “what’s our plan for when this model changes or deprecates?”

PS: AI-generated image - has a few errors.

Things I Learned - 03 May 2026

Sun, 03 May 2026 00:00:00 +0000

This week, I learned:

LiteParse is a PDF to text library that you can run via npx --package=@llamaindex/liteparse lit parse document.pdf. Simon Willison
Always add indecisiveness, inaction, “other”, “not applicable”, etc. as an option to LLMs. They are trained for decisive responses and pattern matching, so we need to guide the the other way. Martin Fowler
GPT 5.5 is priced twice that of GPT 5.4. No wonder my Codex usage is much higher than last month. Simon Willison. I am better off sticking to medium effort instead of the xhigh I usually use - it may not be required. OpenAI
“… the eigenquestion is the question where, if answered, it likely answers the subsequent questions as well.” Shishir Mehrotra & Matt Hudson
Claude Code stores the logged in OAuth token at ~/.claude/.credentials.json. We can use that to fetch https://api.anthropic.com/api/oauth/usage and retrieve Claude usage and reset times. uvx ccusage does this automatically, but I prefer my own script.
Ontology matters in the AI era. But some stuff matters more, and some less.
- 🟢 MORE: Definitions: what “customer” means
- 🟢 MORE: Constraints: e.g. “don’t reclassify loans”
- 🟢 MORE: Interactions: how to verify, coordinate, delegate, …
- 🔴 LESS: Creating ontologies: agents can do that.
- 🔴 LESS: Completeness and rigor: agents tolerate uncertainty.
- 🔴 LESS: Proprietary: agents can reverse-engineer.
There are several industries / markets that MBA case studies rarely cover (ChatGPT): Kirana stores; Care (child care, elder care, domestic work); Faith (finance, food, media, education); Remittances; Gambling (lottery, sports betting, gacha); Scams & organized fraud; Counterfeiting; …

When LLM prices fall 10x every year

Fri, 20 Feb 2026 09:25:45 +0800

In Feb 2024, Claude 3 Opus was the best model, at $15/MTok.
In Jul 2024, GPT 4o Mini reached that quality at 10% of the price.
In Dec 2024, DeepSeek v3 reached that quality at 1% of the price.

Video

See the interactive version

If the price continues to fall 10x every 11-12 months or so (and it has been), then in a year, a Claude 4.6 Opus like model will cost 1/10th of the $5/MTok today, and in 2 years, 1/100th of that.

(We’ll be using better models, of course.)

But 2 years isn’t far away. If Opus 4.6 were 100x cheaper, I could do 100x of what I could do with it today. What would we do with it?

If we assume that they’ll become 100x faster as well (and that’s an important assumption), and the reliability will continue to improve, then:

LLM LSPs. Language servers could be LLMs. Hover over a squiggly line to understand a bug it spotted, right click and fix. Move on.
LLM pre-commit hooks. Write docs, write and run tests, refactor - automatically before you commit.
Continous refactoring. LLMs auto-refactor the code, run tests, and commit better code.
Auto-fix from logs. Log analysis -> Test case -> Fix -> Deployment can be automated.
Pick best option. LLMs generate 30 diverse options for each task, test all, and pick the best. E.g. What’s the best language / framework for this? What’s the better visual design? What should I build?
Live docs. LLMs auto-update docs every commit.
Adversarial workflows. LLMs continously run adversarial test cases to break the code, and fix it.
Build & discard, don’t buy. Most tools are easier to create than purchase. They’re also easier to throw away. To hell with code quality!

The Non-Obvious Impact of Reasoning Defaults

Sun, 05 Oct 2025 02:31:58 +0000

Yesterday, I discovered how much reasoning improves model quality.

My Tools in Data Science assignment asks students to draft an llms.txt file for ipify and auto-checks with GPT-5 Nano - a fast, cheap reasoning model.

I set reasoning_effort to minimal and ran this checklist:

1. Starts with "# ipify" and explains ipify.
2. Markdown sections on API access, support (e.g. GitHub, libraries).
3. Covers API endpoints (IPv4, IPv6, universal) and formats (text, JSON, JSONP).
4. Mentions free, no-auth usage, availability, open-source, safeguards.
5. Has maintenance metadata (e.g. "Last updated: <Month YYYY>").
6. Mentions robots.txt alignment. Stay concise (no filler, <= ~15 links).

If even one checklist item is missing or wrong, fail it.

Respond with EXACTLY one line:
PASS - <brief justification>
or
FAIL - <brief explanation of the first failed item>.

With a perfect llms.txt, it claimed “Metadata section is missing” and “JSONP not mentioned” – though both were present.

With an llms.txt without a metadata or API section, it sometimes marked it as correct!

This surprised me. gpt-5-nano doesn’t usually make such basic mistakes.

Then I realized: reasoning_effort defaults to medium.

When I set reasoning effort back to medium, it added ~5 seconds and ~3,000 reasoning tokens but evaluates correctly.

GPT-5 Nano High is a frontier model on my LLM Pricing Chart, i.e. there’s no cheaper model for that quality. But this conflates GPT-5 Nano cost with GPT-5 Nano High quality. I don’t yet know how to estimate and compare costs of reasoning models.

My takeaways:

My “quantitative” evaluation of frontier models on the LLM Pricing is flawed and needs rethinking.
Reasoning can make smaller models much more powerful. Appreciate it and use reasoning more.
Reasoning models cost far more than non-reasoning models. Benchmark actual costs from usage.

Reasoning is a cheap way to buy accuracy. Just don’t forget you’re paying for it.

Also: beware default settings! Explicit is better than implicit.

Sat, 24 May 2025 07:58:38 +0000

How much does an LLM charge per hour for its services?

If we multiple the Cost Per Output Token with Tokens Per Second, we can get the cost for what an LLM produces in Dollars Per Hour. (We’re ignoring the input cost, but it’s not the main driver of time.)

Over time, different models have been released at different billing rates.

New powerful models like O3 cost ~$7/hr – Poland’s minimum wage rate.
Gemini 2.5 Pro costs ~$12/hr – France’s minimum wage rate.
The latest Claude 4 Sonnet costs ~$2/hr – India’s minimum wage rate.

Once a language model’s run-time cost drops below the local minimum wage, the “offshoring” advantage disappears. AI becomes the cheapest employee in every country at once.

Paradoxically, workers in countries with strong labor protections, unions, and higher wages (like Germany and France) may be safer from AI displacement.

But countries whose economies depend on wage arbitrage face economic disruption.

Visualization: https://public.flourish.studio/visualisation/23384855/
Analysis: https://chatgpt.com/share/68317a06-0cac-800c-ad6f-13646ceb489f
Source code: https://github.com/sanand0/llmpricing/blob/9b66480458b712de52ea14502a64bf4c3b6a98ed/billing.py

Wage Rates of Nations and LLMs

Sat, 24 May 2025 07:47:01 +0000

How much does an LLM charge per hour for its services?

If we multiple the Cost Per Output Token with Tokens Per Second, we can get the cost for what an LLM produces in Dollars Per Hour. (We're ignoring the input cost, but it's not the main driver of time.)

Over time, different models have been released at different billing rates. Most new powerful models like O3 or Gemini 2.5 Pro cost ~$7 - $11 per hr.

To get a sense of this, let's look at wage rates across countries and industries:

Rate ($/hr)	Countries (avg hourly wage)	Models
0–2	Bangladesh ($1.42/hr), Pakistan ($1.65/hr), Vietnam ($0.94/hr)	devstral-small ($0.01/hr), gemini-2.5-flash-preview ($1.50/hr)
2–5	Brazil ($3.09/hr), Mexico manufacturing ($4.90/hr)	claude-sonnet-4 ($2.23/hr), codex-mini ($2.54/hr)
5–10	India ($5.03/hr), South Africa avg ($9.38/hr), Poland min wage ($7.35/hr)	o3 ($7.16/hr), claude-opus-4 ($8.67/hr)
10–15	Germany ($12.93/hr), France ($12.41/hr), UK ($14.43/hr)	gemini-2.5-pro-preview ($11.89/hr), gpt-4.5-preview ($13.10/hr)
15–20	Spain ($15.87/hr), Italy ($16.80/hr), Japan ($17.98/hr)	No recent models in this range

Workers in Europe and Japan are already more expensive than the more expensive models, at $12+ per hour. India, Brazil, Mexico etc. are more expensive than most of the average models.

Once a language model’s run-time cost drops below the local minimum wage, the “offshoring” advantage disappears. AI becomes the cheapest employee in every country at once. Countries whose economies depend on being the "cheaper alternative" for labor-intensive work face potential economic disruption.

Paradoxically, workers in countries with strong labor protections, unions, and higher wages (like Germany and France) may paradoxically be safer from AI displacement.

Source code: sanand0/llmpricing
Analysis: ChatGPT

Comments

Soumendra Kumar Sahoo 24 May 2025 1:44 pm: The white collar workers jobs are in danger compared to blue collar ones.

Fri, 18 Apr 2025 13:10:15 +0000

With the Gemini 2.5 Flash release, Google envelopes the entire cost-quality frontier of LLMs. In other words, at any cost or quality level, today, the best model to use according to the LM Arena score is a Gemini model.

Results for O3, O4 Mini, and GPT 4.1 are not yet on LM Arena. But until then, #Google dominates. Nice work!

Link: https://sanand0.github.io/llmpricing/

Sat, 10 Aug 2024 10:59:35 +0000

Fascinating to see the how LLM cost-quality frontier moves. Recent fights were mostly on cost.

Yesterday, #OpenAI halved the GPT-4o cost. At $2.5/MTok (and with GPT-4o-min at 15 cents/MTok), the best and cheapest models are back with OpenAI, IMHO.

Sigh, time to move all our stuff back from #Anthropic. For now…

https://gramener.com/llmpricing/