All articles

LLM Economics

Token math, model selection, and the true unit cost of every generation.

Distillation Just Got 15x Cheaper. Here Is What That Does to Your AI Cost Model
LLM Economics
Aug 11, 20266 min read

Distillation Just Got 15x Cheaper. Here Is What That Does to Your AI Cost Model

A new memory-efficient training method cut peak VRAM in knowledge distillation from 85.2 GiB to 5.45 GiB at 32K context, which moves the volume at which running your own small model beats paying per token.

DeepSeek's API Price Hike: Why Cheap Tokens Were Never a Strategy
LLM Economics
Aug 10, 20265 min read

DeepSeek's API Price Hike: Why Cheap Tokens Were Never a Strategy

DeepSeek announced on August 6, 2026 that it will raise API prices by a relatively large margin, and the smartest response for AI builders is to model their exposure now, before the number lands.

LLM API Pricing Comparison 2026: GPT, Claude, Gemini and DeepSeek by Blended Cost Per Million Tokens
LLM Economics
Aug 8, 20266 min read

LLM API Pricing Comparison 2026: GPT, Claude, Gemini and DeepSeek by Blended Cost Per Million Tokens

The cheapest LLM API in August 2026 is DeepSeek V4-Flash at $0.14/M input and $0.28/M output, but the number that decides your gross margin is the blended rate for your own input:output mix, not any provider's headline input price.

Open-Weight vs Frontier LLMs: How to Model the Real Cost Difference
LLM Economics
Aug 6, 20265 min read

Open-Weight vs Frontier LLMs: How to Model the Real Cost Difference

Open-weight models are close enough to frontier quality that the switch is now a pricing decision, but the per-token sticker price hides the serving, safety and compliance work the API price was quietly covering.

Inference Engineering Is a Gross Margin Lever, Not an Infra Detail
LLM Economics
Aug 3, 20265 min read

Inference Engineering Is a Gross Margin Lever, Not an Infra Detail

Inference engineering is the discipline of turning model weights into a fast, affordable production API, and the throughput gains it produces land in someone's gross margin: yours if you serve the model, your provider's if you buy tokens.

LLM Economics
Aug 3, 20265 min read

Open-Weight Models and Your LLM Cost Per Token: What DeepSeek V4 Flash Actually Changes

An open-weight model landing within a few points of the frontier does not automatically cut your AI bill: your real cost per token is set by how many tokens you consume, how fast you can serve them, and whether you own the hardware at all.

LLM Prices Fell 13x in Four Months: What That Actually Does to Your AI Margins
LLM Economics
Aug 2, 20266 min read

LLM Prices Fell 13x in Four Months: What That Actually Does to Your AI Margins

Frontier-level intelligence now costs roughly one-thirteenth what it did four months ago, but a 13x cut in list price only becomes a 13x cut in your COGS if you re-route workloads and hold token consumption flat.

Flat AI Pricing Is a Subsidy: One User Burned $30,983 of Tokens on a $200 Plan
LLM Economics
Aug 1, 20266 min read

Flat AI Pricing Is a Subsidy: One User Burned $30,983 of Tokens on a $200 Plan

Flat AI subscriptions only work while heavy users stay rare: one developer consuming roughly $30,983 of token value on a $200 per month plan is about a 155x gap between what was paid and what it cost to serve.

OpenAI's GPT-5.6 Price Cut: What Luna at $0.20 and Terra at $2 Do to Your Margins
LLM Economics
Jul 31, 20266 min read

OpenAI's GPT-5.6 Price Cut: What Luna at $0.20 and Terra at $2 Do to Your Margins

Starting July 30, 2026, GPT-5.6 Luna costs $0.20 per 1M input tokens and $1.20 per 1M output tokens (an 80% cut), and Terra costs $2 and $12 (a 20% cut), which means the biggest saving for most teams comes from re-routing workloads rather than from the discount itself.

Cheaper Tokens Do Not Equal Better Margins: What GPT-5.6's Efficiency Push Means for Your AI Unit Economics
LLM Economics
Jul 30, 20265 min read

Cheaper Tokens Do Not Equal Better Margins: What GPT-5.6's Efficiency Push Means for Your AI Unit Economics

Provider efficiency gains only reach your P&L if your own architecture captures them: your prompt-cache hit rate, your agent loop length, and your model tiering decide whether a 20% serving-cost cut becomes 20% of margin or nothing at all.

Claude Opus 5 vs Fable 5: What the Half-Price Flagship Means for Your LLM Cost Model
LLM Economics
Jul 28, 20264 min read

Claude Opus 5 vs Fable 5: What the Half-Price Flagship Means for Your LLM Cost Model

Anthropic's Claude Opus 5 lands within a few benchmark points of Fable 5 at roughly half the price, and independent testing shows a 20% lower cost per task, a shift worth rerunning your margin math over.

Why the GPU-Hour Sticker Price Isn't What You'll Pay for AI Inference at Scale
LLM Economics
Jul 27, 20265 min read

Why the GPU-Hour Sticker Price Isn't What You'll Pay for AI Inference at Scale

Headline GPU and LLM token prices describe what one unit costs, not what a usable amount of compute costs once you need it delivered together, on time, and at your scale.

The Margin Memo

Pricing math, in your inbox.

One short note a week on AI pricing, token economics, and margin. No spam, unsubscribe anytime.